[llvm] [AMDGPU] Fix DS result latency for inline asm (PR #229538)

Alan Li via llvm-commits llvm-commits at lists.llvm.org
Tue Oct 6 14:51:42 PDT 2026


https://github.com/lialan updated https://github.com/llvm/llvm-project/pull/229538

>From 646ef0dc9679babc4001c7fc5b5466357444e1da Mon Sep 17 00:00:00 2001
From: Alan Li <me at alanli.org>
Date: Mon, 5 Oct 2026 15:29:58 -0500
Subject: [PATCH] [AMDGPU] Restore producer latency for inline asm

Explicit physical-register inline-asm inputs can receive zero latency
because the generic scheduling DAG classifies variable operands as pseudo
uses. Restore modeled producer latency at explicit INLINEASM inputs through
GCNSubtarget::adjustSchedDependency, independently of the producer's
instruction class, address space, scheduling strategy, or register
allocation phase. Virtual-register edges already compute operand latency.

Preserve the implicit-operand exclusions for register bookkeeping. Cover
DS reads, returning LDS and GDS atomics, fixed-GDS results, VALU, VMEM, and
MFMA producers, as well as tied and untied inputs and scheduling order.
Update the affected instruction-order and wait-insertion checks.
---
 llvm/lib/Target/AMDGPU/GCNSubtarget.cpp       |   9 +
 .../test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll |  68 ++---
 .../CodeGen/AMDGPU/a-v-global-atomicrmw.ll    |   4 +-
 llvm/test/CodeGen/AMDGPU/agpr-remat.ll        |   2 +-
 .../CodeGen/AMDGPU/captured-frame-index.ll    |   6 +-
 llvm/test/CodeGen/AMDGPU/fneg-combines.ll     |   4 +-
 llvm/test/CodeGen/AMDGPU/gds-allocation.ll    |   2 +-
 .../gfx-callable-preserved-registers.ll       |  22 +-
 .../AMDGPU/inline-asm-vgpr-sgpr-copy.ll       |   4 +-
 .../AMDGPU/inlineasm-ds-return-latency.mir    |  47 ++++
 .../AMDGPU/inlineasm-producer-latency.mir     |  47 ++++
 .../CodeGen/AMDGPU/insert_vector_elt.v2i16.ll |   6 +-
 .../AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll     |  24 +-
 .../AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll        |   4 +-
 .../AMDGPU/llvm.amdgcn.readfirstlane.ll       |  84 +++---
 .../materialize-frame-index-sgpr.gfx10.ll     |  34 +--
 .../AMDGPU/materialize-frame-index-sgpr.ll    |   2 +-
 .../AMDGPU/postra-coexec-inlineasm-flags.mir  |  49 ++++
 .../postra-coexec-inlineasm-latency.mir       |  93 +++++++
 .../AMDGPU/preserve-wwm-copy-dst-reg.ll       |  72 ++---
 .../scc-clobbered-sgpr-to-vmem-spill.ll       |  24 +-
 .../AMDGPU/shufflevector.v2i64.v8i64.ll       | 262 +++++++++---------
 llvm/test/CodeGen/AMDGPU/spill-agpr.ll        |  14 +-
 .../unspill-vgpr-after-rewrite-vgpr-mfma.ll   |   4 +-
 .../CodeGen/AMDGPU/wmma-scale-src-perm.ll     |   6 +-
 25 files changed, 569 insertions(+), 324 deletions(-)
 create mode 100644 llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir
 create mode 100644 llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir
 create mode 100644 llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir
 create mode 100644 llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir

diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp b/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
index 31dd8c8be27818..560ca504639441 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
@@ -850,6 +850,15 @@ void GCNSubtarget::adjustSchedDependency(
     return; // This is not a data dependency anymore.
   }
 
+  // Inline asm's variable explicit inputs are classified as pseudo uses by
+  // the generic physical-register DAG builder and receive zero latency.
+  // Restore the modeled producer latency for explicit inline-asm inputs.
+  if (UseI->getOpcode() == TargetOpcode::INLINEASM &&
+      !UseI->getOperand(UseOpIdx).isImplicit() &&
+      !DefI->getOperand(DefOpIdx).isImplicit())
+    Dep.setLatency(
+        SchedModel->computeOperandLatency(DefI, DefOpIdx, UseI, UseOpIdx));
+
   if (DefI->isBundle()) {
     const SIRegisterInfo *TRI = getRegisterInfo();
     auto Reg = Dep.getReg();
diff --git a/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll b/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll
index fcb17b5865c32d..1e319e5850ea86 100644
--- a/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll
+++ b/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll
@@ -536,8 +536,8 @@ define void @flat_atomic_xchg_i32_ret_av_av_no_agprs(ptr %ptr) #0 {
 ; GFX950-NEXT:    scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v23, a27 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v22, a28 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v21, a29 ; Reload Reuse
@@ -4271,8 +4271,8 @@ define void @flat_atomic_xor_i32_ret_av_av_no_agprs(ptr %ptr) #0 {
 ; GFX950-NEXT:    scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v23, a27 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v22, a28 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v21, a29 ; Reload Reuse
@@ -14634,12 +14634,12 @@ define void @flat_atomic_add_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v4, v2, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    s_waitcnt vmcnt(1)
 ; GFX90A-NEXT:    v_add_co_u32_e32 v0, vcc, v3, v0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
 ; GFX90A-NEXT:    v_addc_co_u32_e32 v1, vcc, v4, v1, vcc
-; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v1, v2, s[0:3], 0 offen offset:4
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -14675,9 +14675,9 @@ define void @flat_atomic_add_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    scratch_load_dwordx2 v[2:3], off, s0
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_lshl_add_u64 v[0:1], v[2:3], 0, v[0:1]
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -14797,12 +14797,12 @@ define void @flat_atomic_sub_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v4, v2, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    s_waitcnt vmcnt(1)
 ; GFX90A-NEXT:    v_sub_co_u32_e32 v0, vcc, v3, v0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
 ; GFX90A-NEXT:    v_subb_co_u32_e32 v1, vcc, v4, v1, vcc
-; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v1, v2, s[0:3], 0 offen offset:4
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -14841,8 +14841,8 @@ define void @flat_atomic_sub_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    s_nop 0
 ; GFX950-NEXT:    v_subb_co_u32_e32 v1, vcc, v3, v1, vcc
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -14963,10 +14963,10 @@ define void @flat_atomic_and_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v3, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_load_dword v4, v2, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    s_waitcnt vmcnt(1)
-; GFX90A-NEXT:    v_and_b32_e32 v0, v3, v0
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
+; GFX90A-NEXT:    v_and_b32_e32 v0, v3, v0
 ; GFX90A-NEXT:    v_and_b32_e32 v1, v4, v1
 ; GFX90A-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v1, v2, s[0:3], 0 offen offset:4
@@ -15006,9 +15006,9 @@ define void @flat_atomic_and_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_and_b32_e32 v1, v3, v1
 ; GFX950-NEXT:    v_and_b32_e32 v0, v2, v0
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -15361,10 +15361,10 @@ define void @flat_atomic_or_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v3, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_load_dword v4, v2, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    s_waitcnt vmcnt(1)
-; GFX90A-NEXT:    v_or_b32_e32 v0, v3, v0
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
+; GFX90A-NEXT:    v_or_b32_e32 v0, v3, v0
 ; GFX90A-NEXT:    v_or_b32_e32 v1, v4, v1
 ; GFX90A-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v1, v2, s[0:3], 0 offen offset:4
@@ -15404,9 +15404,9 @@ define void @flat_atomic_or_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_or_b32_e32 v1, v3, v1
 ; GFX950-NEXT:    v_or_b32_e32 v0, v2, v0
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -15526,10 +15526,10 @@ define void @flat_atomic_xor_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v3, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_load_dword v4, v2, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    s_waitcnt vmcnt(1)
-; GFX90A-NEXT:    v_xor_b32_e32 v0, v3, v0
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v3
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v4
+; GFX90A-NEXT:    v_xor_b32_e32 v0, v3, v0
 ; GFX90A-NEXT:    v_xor_b32_e32 v1, v4, v1
 ; GFX90A-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v1, v2, s[0:3], 0 offen offset:4
@@ -15569,9 +15569,9 @@ define void @flat_atomic_xor_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_xor_b32_e32 v1, v3, v1
 ; GFX950-NEXT:    v_xor_b32_e32 v0, v2, v0
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -15695,10 +15695,10 @@ define void @flat_atomic_max_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_cmp_gt_i64_e32 vcc, v[2:3], v[0:1]
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -15865,10 +15865,10 @@ define void @flat_atomic_min_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_cmp_le_i64_e32 vcc, v[2:3], v[0:1]
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -16035,10 +16035,10 @@ define void @flat_atomic_umax_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1]
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -16205,10 +16205,10 @@ define void @flat_atomic_umin_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_cmp_le_u64_e32 vcc, v[2:3], v[0:1]
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -16375,12 +16375,12 @@ define void @flat_atomic_uinc_wrap_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_addc_co_u32_e32 v6, vcc, 0, v3, vcc
 ; GFX90A-NEXT:    v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1]
-; GFX90A-NEXT:    v_cndmask_b32_e32 v1, 0, v5, vcc
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT:    v_cndmask_b32_e32 v1, 0, v5, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, 0, v6, vcc
 ; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen offset:4
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -16552,12 +16552,12 @@ define void @flat_atomic_udec_wrap_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    v_subbrev_co_u32_e32 v6, vcc, 0, v3, vcc
 ; GFX90A-NEXT:    v_cmp_gt_u64_e64 s[4:5], v[2:3], v[0:1]
 ; GFX90A-NEXT:    s_or_b64 vcc, vcc, s[4:5]
-; GFX90A-NEXT:    v_cndmask_b32_e32 v1, v6, v1, vcc
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT:    v_cndmask_b32_e32 v1, v6, v1, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, v5, v0, vcc
 ; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -16600,9 +16600,9 @@ define void @flat_atomic_udec_wrap_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    s_nop 0
 ; GFX950-NEXT:    v_cndmask_b32_e32 v1, v5, v1, vcc
 ; GFX950-NEXT:    v_cndmask_b32_e32 v0, v4, v0, vcc
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s2
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -16738,10 +16738,10 @@ define void @flat_atomic_usub_cond_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    v_cmp_ge_u64_e32 vcc, v[0:1], v[4:5]
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v0
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v1
 ; GFX90A-NEXT:    v_cndmask_b32_e32 v4, v1, v6, vcc
 ; GFX90A-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
 ; GFX90A-NEXT:    buffer_store_dword v4, v2, s[0:3], 0 offen offset:4
-; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v1
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -17236,8 +17236,8 @@ define void @flat_atomic_fadd_f32_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v2, v1, s[0:3], 0 offen
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_add_f32_e32 v0, v2, v0
-; GFX90A-NEXT:    buffer_store_dword v0, v1, s[0:3], 0 offen
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT:    buffer_store_dword v0, v1, s[0:3], 0 offen
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a0
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18066,10 +18066,10 @@ define void @flat_atomic_fadd_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    buffer_load_dword v3, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_add_f64 v[0:1], v[2:3], v[0:1]
-; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT:    buffer_store_dword v0, v4, s[0:3], 0 offen
+; GFX90A-NEXT:    buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18117,9 +18117,9 @@ define void @flat_atomic_fadd_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    scratch_load_dwordx2 v[2:3], off, s0
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_add_f64 v[0:1], v[2:3], v[0:1]
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -18475,10 +18475,10 @@ define void @flat_atomic_fmax_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_max_f64 v[4:5], v[2:3], v[2:3]
 ; GFX90A-NEXT:    v_max_f64 v[0:1], v[4:5], v[0:1]
-; 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:    v_accvgpr_write_b32 a0, v2
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
+; 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:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18516,9 +18516,9 @@ define void @flat_atomic_fmax_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_max_f64 v[4:5], v[2:3], v[2:3]
 ; GFX950-NEXT:    v_max_f64 v[0:1], v[4:5], v[0:1]
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
@@ -18643,10 +18643,10 @@ define void @flat_atomic_fmin_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    v_max_f64 v[4:5], v[2:3], v[2:3]
 ; GFX90A-NEXT:    v_min_f64 v[0:1], v[4:5], v[0:1]
-; 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:    v_accvgpr_write_b32 a0, v2
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
+; 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:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use a[0:1]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18684,9 +18684,9 @@ define void @flat_atomic_fmin_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
 ; GFX950-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-NEXT:    v_max_f64 v[4:5], v[2:3], v[2:3]
 ; GFX950-NEXT:    v_min_f64 v[0:1], v[4:5], v[0:1]
-; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    v_accvgpr_write_b32 a0, v2
 ; GFX950-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT:    scratch_store_dwordx2 off, v[0:1], s0
 ; GFX950-NEXT:    ;;#ASMSTART
 ; GFX950-NEXT:    ; use a[0:1]
 ; GFX950-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll b/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll
index 4983695ac24ee9..a3b3f7e65c6ba8 100644
--- a/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll
+++ b/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll
@@ -536,8 +536,8 @@ define void @global_atomic_xchg_i32_ret_av_av_no_agprs(ptr addrspace(1) %ptr) #0
 ; GFX950-NEXT:    scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v23, a27 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v22, a28 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v21, a29 ; Reload Reuse
@@ -3262,8 +3262,8 @@ define void @global_atomic_xor_i32_ret_av_av_no_agprs(ptr addrspace(1) %ptr) #0
 ; GFX950-NEXT:    scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
 ; GFX950-NEXT:    scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT:    v_accvgpr_read_b32 v19, a31 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v23, a27 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v22, a28 ; Reload Reuse
 ; GFX950-NEXT:    v_accvgpr_read_b32 v21, a29 ; Reload Reuse
diff --git a/llvm/test/CodeGen/AMDGPU/agpr-remat.ll b/llvm/test/CodeGen/AMDGPU/agpr-remat.ll
index ede7ce390e1598..dd7506297fa0e4 100644
--- a/llvm/test/CodeGen/AMDGPU/agpr-remat.ll
+++ b/llvm/test/CodeGen/AMDGPU/agpr-remat.ll
@@ -32,9 +32,9 @@ define void @remat_regcopy_avoids_spill(i32 %v0, i32 %v1, i32 %v2, i32 %v3, i32
 ; GFX908-NEXT:    v_accvgpr_write_b32 a2, v0
 ; GFX908-NEXT:    v_accvgpr_write_b32 a3, v1
 ; GFX908-NEXT:    v_accvgpr_write_b32 a4, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v3
 ; GFX908-NEXT:    v_accvgpr_write_b32 a0, v7
 ; GFX908-NEXT:    v_accvgpr_write_b32 a1, v8
-; GFX908-NEXT:    v_accvgpr_write_b32 a5, v3
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ;;#ASMEND
 ; GFX908-NEXT:    v_accvgpr_write_b32 a2, v4
diff --git a/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll b/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
index 28d650013fcec5..5916b08be4727b 100644
--- a/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
+++ b/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
@@ -273,10 +273,10 @@ define void @func_alloca_offset_use_asm_sgpr() {
 ; GCN: v_lshr_b32_e64 [[V_FI:v[0-9]+]], s32, 6
 ; GCN: s_movk_i32 vcc_lo, 0x4010
 ; GCN: s_add_i32 [[S_FI]], [[S_FI]], 16
+; GCN-NEXT: v_add_i32_e32 [[V_FI]], vcc, vcc_lo, [[V_FI]]
 ; GCN-NEXT: ;;#ASMSTART
 ; GCN-NEXT: ; use [[S_FI]]
 ; GCN-NEXT: ;;#ASMEND
-; GCN-NEXT: v_add_i32_e32 [[V_FI:v[0-9]+]], vcc, vcc_lo, [[V_FI]]
 ; GCN-NEXT: ;;#ASMSTART
 ; GCN-NEXT: ; use [[V_FI]]
 ; GCN-NEXT: ;;#ASMEND
@@ -295,7 +295,7 @@ define void @func_alloca_offset_use_asm_vgpr() {
 ; GCN-NEXT: ; use [[FI0]]
 ; GCN-NEXT: ;;#ASMEND
 ; GCN: s_movk_i32 [[FI1:s[0-9]+]], 0x4010
-; GCN-NEXT: ;;#ASMSTART
+; GCN: ;;#ASMSTART
 ; GCN-NEXT: ; use [[FI1]]
 ; GCN-NEXT: ;;#ASMEND
 define amdgpu_kernel void @kernel_alloca_offset_use_asm_sgpr() {
@@ -314,7 +314,7 @@ define amdgpu_kernel void @kernel_alloca_offset_use_asm_sgpr() {
 ; GCN-NEXT: ;;#ASMEND
 
 ; GCN: v_mov_b32_e32 v0, 0x4010
-; GCN-NEXT: ;;#ASMSTART
+; GCN: ;;#ASMSTART
 ; GCN-NEXT: ; use v0
 ; GCN-NEXT: ;;#ASMEND
 define amdgpu_kernel void @kernel_alloca_offset_use_asm_vgpr() {
diff --git a/llvm/test/CodeGen/AMDGPU/fneg-combines.ll b/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
index 5395bb784fc935..21de0e507c162d 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
@@ -7203,11 +7203,11 @@ define amdgpu_kernel void @v_fneg_inlineasm_multi_use_src_f32(ptr addrspace(1) %
 ; SI-NEXT:    v_mov_b32_e32 v3, s5
 ; SI-NEXT:    v_add_i32_e32 v2, vcc, s4, v4
 ; SI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
-; SI-NEXT:    v_mov_b32_e32 v5, s7
 ; SI-NEXT:    flat_load_dword v6, v[0:1] glc
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    flat_load_dword v2, v[2:3] glc
 ; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_mov_b32_e32 v5, s7
 ; SI-NEXT:    v_add_i32_e32 v0, vcc, s6, v4
 ; SI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v5, vcc
 ; SI-NEXT:    flat_load_dword v0, v[0:1] glc
@@ -7235,11 +7235,11 @@ define amdgpu_kernel void @v_fneg_inlineasm_multi_use_src_f32(ptr addrspace(1) %
 ; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    v_add_u32_e32 v2, vcc, s4, v4
 ; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
-; VI-NEXT:    v_mov_b32_e32 v5, s7
 ; VI-NEXT:    flat_load_dword v6, v[0:1] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    flat_load_dword v2, v[2:3] glc
 ; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v5, s7
 ; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v4
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v5, vcc
 ; VI-NEXT:    flat_load_dword v0, v[0:1] glc
diff --git a/llvm/test/CodeGen/AMDGPU/gds-allocation.ll b/llvm/test/CodeGen/AMDGPU/gds-allocation.ll
index 5c92976734c8be..ca1567b1e724d9 100644
--- a/llvm/test/CodeGen/AMDGPU/gds-allocation.ll
+++ b/llvm/test/CodeGen/AMDGPU/gds-allocation.ll
@@ -107,9 +107,9 @@ define amdgpu_kernel void @gds_extern_align(ptr addrspace(1) %out, ptr addrspace
 ; GCN-LABEL: gds_extern_align:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dword s0, s[4:5], 0x8
+; GCN-NEXT:    s_movk_i32 s1, 0x400
 ; GCN-NEXT:    v_mov_b32_e32 v0, 5
 ; GCN-NEXT:    s_movk_i32 m0, 0x401
-; GCN-NEXT:    s_movk_i32 s1, 0x400
 ; GCN-NEXT:    ;;#ASMSTART
 ; GCN-NEXT:    ; use s1
 ; GCN-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
index 7d028b383818de..ad0ac3c6467da2 100644
--- a/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
@@ -258,12 +258,12 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_s31(ptr addrspace(1)
 ; GFX10-NEXT:    s_mov_b32 s4, s31
 ; GFX10-NEXT:    s_swappc_b64 s[30:31], s[34:35]
 ; GFX10-NEXT:    s_mov_b32 s31, s4
+; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
 ; GFX10-NEXT:    ;;#ASMSTART
 ; GFX10-NEXT:    ; use s31
 ; GFX10-NEXT:    ;;#ASMEND
 ; GFX10-NEXT:    v_readlane_b32 s30, v40, 1
 ; GFX10-NEXT:    v_readlane_b32 s31, v40, 2
-; GFX10-NEXT:    v_readlane_b32 s4, v40, 0
 ; GFX10-NEXT:    s_mov_b32 s32, s33
 ; GFX10-NEXT:    v_readlane_b32 s34, v40, 3
 ; GFX10-NEXT:    s_or_saveexec_b32 s35, -1
@@ -295,12 +295,12 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_s31(ptr addrspace(1)
 ; GFX11-NEXT:    s_mov_b32 s4, s31
 ; GFX11-NEXT:    s_swappc_b64 s[30:31], s[0:1]
 ; GFX11-NEXT:    s_mov_b32 s31, s4
+; GFX11-NEXT:    v_readlane_b32 s4, v40, 0
 ; GFX11-NEXT:    ;;#ASMSTART
 ; GFX11-NEXT:    ; use s31
 ; GFX11-NEXT:    ;;#ASMEND
 ; GFX11-NEXT:    v_readlane_b32 s30, v40, 1
 ; GFX11-NEXT:    v_readlane_b32 s31, v40, 2
-; GFX11-NEXT:    v_readlane_b32 s4, v40, 0
 ; GFX11-NEXT:    s_mov_b32 s32, s33
 ; GFX11-NEXT:    v_readlane_b32 s0, v40, 3
 ; GFX11-NEXT:    s_or_saveexec_b32 s1, -1
@@ -374,14 +374,14 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_v31(ptr addrspace(1)
 ; GFX10-NEXT:    v_mov_b32_e32 v40, v31
 ; GFX10-NEXT:    s_swappc_b64 s[30:31], s[34:35]
 ; GFX10-NEXT:    v_mov_b32_e32 v31, v40
-; GFX10-NEXT:    ;;#ASMSTART
-; GFX10-NEXT:    ; use v31
-; GFX10-NEXT:    ;;#ASMEND
-; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
 ; GFX10-NEXT:    v_readlane_b32 s30, v41, 0
 ; GFX10-NEXT:    v_readlane_b32 s31, v41, 1
 ; GFX10-NEXT:    s_mov_b32 s32, s33
 ; GFX10-NEXT:    v_readlane_b32 s34, v41, 2
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    ; use v31
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
 ; GFX10-NEXT:    s_or_saveexec_b32 s35, -1
 ; GFX10-NEXT:    buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
 ; GFX10-NEXT:    s_waitcnt_depctr depctr_vm_vsrc(0)
@@ -412,14 +412,14 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_v31(ptr addrspace(1)
 ; GFX11-NEXT:    s_swappc_b64 s[30:31], s[0:1]
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_mov_b32_e32 v31, v40
-; GFX11-NEXT:    ;;#ASMSTART
-; GFX11-NEXT:    ; use v31
-; GFX11-NEXT:    ;;#ASMEND
-; GFX11-NEXT:    scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
 ; GFX11-NEXT:    v_readlane_b32 s30, v41, 0
 ; GFX11-NEXT:    v_readlane_b32 s31, v41, 1
 ; GFX11-NEXT:    s_mov_b32 s32, s33
 ; GFX11-NEXT:    v_readlane_b32 s0, v41, 2
+; GFX11-NEXT:    ;;#ASMSTART
+; GFX11-NEXT:    ; use v31
+; GFX11-NEXT:    ;;#ASMEND
+; GFX11-NEXT:    scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
 ; GFX11-NEXT:    s_or_saveexec_b32 s1, -1
 ; GFX11-NEXT:    scratch_load_b32 v41, off, s33 offset:4 ; 4-byte Folded Reload
 ; GFX11-NEXT:    s_mov_b32 exec_lo, s1
@@ -644,8 +644,8 @@ define amdgpu_gfx void @test_call_void_func_void_preserves_s34(ptr addrspace(1)
 ; GFX11-NEXT:    ;;#ASMEND
 ; GFX11-NEXT:    s_mov_b32 s4, s34
 ; GFX11-NEXT:    s_swappc_b64 s[30:31], s[0:1]
-; GFX11-NEXT:    v_readlane_b32 s30, v40, 1
 ; GFX11-NEXT:    s_mov_b32 s34, s4
+; GFX11-NEXT:    v_readlane_b32 s30, v40, 1
 ; GFX11-NEXT:    ;;#ASMSTART
 ; GFX11-NEXT:    ; use s34
 ; GFX11-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll b/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll
index 73b2ce3a50774e..9fcee491b8a31d 100644
--- a/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll
+++ b/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll
@@ -10,11 +10,11 @@ define <4 x float> @test_sgpr_constraint_bug(ptr addrspace(5) %buf_desc_ptr) {
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; CHECK-NEXT:    scratch_load_dwordx4 v[4:7], v0, off
 ; CHECK-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-NEXT:    s_mov_b64 s[4:5], exec
+; CHECK-NEXT:    v_mov_b32_e32 v8, 1
 ; CHECK-NEXT:    v_mov_b32_e32 v1, v0
 ; CHECK-NEXT:    v_mov_b32_e32 v2, v0
 ; CHECK-NEXT:    v_mov_b32_e32 v3, v0
-; CHECK-NEXT:    s_mov_b64 s[4:5], exec
-; CHECK-NEXT:    v_mov_b32_e32 v8, 1
 ; CHECK-NEXT:    s_waitcnt vmcnt(0)
 ; CHECK-NEXT:    v_readfirstlane_b32 s3, v7
 ; CHECK-NEXT:    v_readfirstlane_b32 s2, v6
diff --git a/llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir b/llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir
new file mode 100644
index 00000000000000..39dfe2f30554a9
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir
@@ -0,0 +1,47 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -run-pass=postmisched -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Explicit inline-asm inputs retain native DS return latency, including LDS
+# atomics, configurable GDS operations, and fixed-GDS results.
+
+---
+name: atomic_lds_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: atomic_lds_result:%bb.0
+  ; CHECK: SU(0): {{.*}}DS_ADD_RTN_U32_gfx9
+  ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $exec
+    $vgpr2 = DS_ADD_RTN_U32_gfx9 $vgpr0, $vgpr1, 0, 0, implicit $exec :: (load store (s32), addrspace 3)
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
+
+---
+name: atomic_configurable_gds_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: atomic_configurable_gds_result:%bb.0
+  ; CHECK: SU(0): {{.*}}DS_ADD_RTN_U32
+  ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $m0, $exec
+    $vgpr2 = DS_ADD_RTN_U32 $vgpr0, $vgpr1, 0, 1, implicit $m0, implicit $exec :: (load store (s32), addrspace 2)
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
+
+---
+name: fixed_gds_ordered_count_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: fixed_gds_ordered_count_result:%bb.0
+  ; CHECK: SU(0): {{.*}}DS_ORDERED_COUNT
+  ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $m0, $exec
+    $vgpr2 = DS_ORDERED_COUNT $vgpr0, 772, implicit $m0, implicit $exec :: (load store (s32), addrspace 2)
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir b/llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir
new file mode 100644
index 00000000000000..aa12bbd9d76beb
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir
@@ -0,0 +1,47 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Explicit inline-asm inputs retain modeled producer latency for non-DS
+# instructions, including arithmetic, global loads, and matrix instructions.
+
+---
+name: valu_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: valu_result:%bb.0
+  ; CHECK: SU(0): {{.*}}V_ADD_U32_e32
+  ; CHECK: SU(1): Data Latency=1 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $exec
+    $vgpr2 = V_ADD_U32_e32 1, $vgpr0, implicit $exec
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
+
+---
+name: vmem_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: vmem_result:%bb.0
+  ; CHECK: SU(0): {{.*}}GLOBAL_LOAD_DWORD
+  ; CHECK: SU(1): Data Latency=80 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0_vgpr1, $exec
+    $vgpr2 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s32), addrspace 1)
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
+
+---
+name: mfma_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: mfma_result:%bb.0
+  ; CHECK: SU(0): {{.*}}V_MFMA_F32_16X16X16F16_vgprcd_e64
+  ; CHECK: SU(1): Data Latency=4 Reg=$vgpr4_vgpr5_vgpr6_vgpr7
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr8_vgpr9_vgpr10_vgpr11, $mode, $exec
+    $vgpr4_vgpr5_vgpr6_vgpr7 = V_MFMA_F32_16X16X16F16_vgprcd_e64 $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr8_vgpr9_vgpr10_vgpr11, 0, 0, 0, implicit $mode, implicit $exec
+    INLINEASM &"", attdialect, regdef:VReg_128_Align2, def $vgpr4_vgpr5_vgpr6_vgpr7, reguse tiedto:$0, $vgpr4_vgpr5_vgpr6_vgpr7(tied-def 3)
+    S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
index 651640d514e436..8ffeaa913d8c2d 100644
--- a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
@@ -155,8 +155,8 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_multi_use_hi_reg(ptr addrspac
 ; VI-NEXT:    v_mov_b32_e32 v1, s1
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s2
-; VI-NEXT:    v_perm_b32 v2, s4, v3, v2
 ; VI-NEXT:    s_lshr_b32 s0, s2, 16
+; VI-NEXT:    v_perm_b32 v2, s4, v3, v2
 ; VI-NEXT:    flat_store_dword v[0:1], v2
 ; VI-NEXT:    ;;#ASMSTART
 ; VI-NEXT:    ; use s0
@@ -178,8 +178,8 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_multi_use_hi_reg(ptr addrspac
 ; CI-NEXT:    s_waitcnt lgkmcnt(0)
 ; CI-NEXT:    s_and_b32 s1, s2, 0xffff0000
 ; CI-NEXT:    s_or_b32 s0, s0, s1
-; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    s_lshr_b32 s2, s2, 16
+; CI-NEXT:    v_mov_b32_e32 v2, s0
 ; CI-NEXT:    flat_store_dword v[0:1], v2
 ; CI-NEXT:    ;;#ASMSTART
 ; CI-NEXT:    ; use s2
@@ -409,8 +409,8 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_reghi_both_multi_use_1(ptr ad
 ; VI-NEXT:    s_lshr_b32 s0, s4, 16
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    v_mov_b32_e32 v3, s2
-; VI-NEXT:    v_perm_b32 v2, s4, v3, v2
 ; VI-NEXT:    s_lshr_b32 s1, s2, 16
+; VI-NEXT:    v_perm_b32 v2, s4, v3, v2
 ; VI-NEXT:    flat_store_dword v[0:1], v2
 ; VI-NEXT:    ;;#ASMSTART
 ; VI-NEXT:    ; use s0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
index ffc2e2d0bcb911..14c00e9ebda38f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
@@ -19,11 +19,11 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v0, s2
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v1, s3
 ; SI-SDAG-NEXT:    s_movk_i32 s0, 0x64
-; SI-SDAG-NEXT:    s_mov_b32 s5, s1
 ; SI-SDAG-NEXT:    ;;#ASMSTART
 ; SI-SDAG-NEXT:    v_lsrlrev_b64 v[4:5], v[0:1], 1
 ; SI-SDAG-NEXT:    ;;#ASMEND
 ; SI-SDAG-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[4:5], s0, v[2:3]
+; SI-SDAG-NEXT:    s_mov_b32 s5, s1
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v5, v1
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v4, v0
 ; SI-SDAG-NEXT:    ;;#ASMSTART
@@ -46,9 +46,9 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
 ; SI-GISEL-NEXT:    ;;#ASMEND
 ; SI-GISEL-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[4:5], v6, v[2:3]
 ; SI-GISEL-NEXT:    s_mov_b32 s2, -1
-; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v5, v1
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-GISEL-NEXT:    ;;#ASMSTART
 ; SI-GISEL-NEXT:    ;; force constraint
 ; SI-GISEL-NEXT:    ;;#ASMEND
@@ -64,12 +64,12 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
 ; VI-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v0, s2
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v1, s3
-; VI-SDAG-NEXT:    v_mov_b32_e32 v2, s0
-; VI-SDAG-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-SDAG-NEXT:    ;;#ASMSTART
 ; VI-SDAG-NEXT:    v_lsrlrev_b64 v[4:5], v[0:1], 1
 ; VI-SDAG-NEXT:    ;;#ASMEND
 ; VI-SDAG-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[4:5], s4, v[6:7]
+; VI-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; VI-SDAG-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v5, v1
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v4, v0
 ; VI-SDAG-NEXT:    ;;#ASMSTART
@@ -91,10 +91,10 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
 ; VI-GISEL-NEXT:    v_lsrlrev_b64 v[4:5], v[0:1], 1
 ; VI-GISEL-NEXT:    ;;#ASMEND
 ; VI-GISEL-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[4:5], v6, v[2:3]
-; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s1
-; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v5, v1
 ; VI-GISEL-NEXT:    v_mov_b32_e32 v4, v0
+; VI-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; VI-GISEL-NEXT:    v_mov_b32_e32 v2, s0
 ; VI-GISEL-NEXT:    ;;#ASMSTART
 ; VI-GISEL-NEXT:    ;; force constraint
 ; VI-GISEL-NEXT:    ;;#ASMEND
@@ -127,12 +127,12 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8_non_immediate(ptr addrspace(1) %out
 ; SI-SDAG-NEXT:    ;;#ASMEND
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SI-SDAG-NEXT:    s_mov_b32 s4, s0
-; SI-SDAG-NEXT:    s_mov_b32 s5, s1
 ; SI-SDAG-NEXT:    ;;#ASMSTART
 ; SI-SDAG-NEXT:    v_lshlrev_b64 v[6:7], v[0:1], 1
 ; SI-SDAG-NEXT:    ;;#ASMEND
 ; SI-SDAG-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[2:3], v4, v[6:7]
+; SI-SDAG-NEXT:    s_mov_b32 s4, s0
+; SI-SDAG-NEXT:    s_mov_b32 s5, s1
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v3, v1
 ; SI-SDAG-NEXT:    v_mov_b32_e32 v2, v0
 ; SI-SDAG-NEXT:    ;;#ASMSTART
@@ -162,10 +162,10 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8_non_immediate(ptr addrspace(1) %out
 ; SI-GISEL-NEXT:    v_lshlrev_b64 v[6:7], v[0:1], 1
 ; SI-GISEL-NEXT:    ;;#ASMEND
 ; SI-GISEL-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[2:3], v4, v[6:7]
-; SI-GISEL-NEXT:    s_mov_b32 s2, -1
-; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v3, v1
 ; SI-GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; SI-GISEL-NEXT:    s_mov_b32 s2, -1
+; SI-GISEL-NEXT:    s_mov_b32 s3, 0xf000
 ; SI-GISEL-NEXT:    ;;#ASMSTART
 ; SI-GISEL-NEXT:    ;; force constraint
 ; SI-GISEL-NEXT:    ;;#ASMEND
@@ -189,12 +189,12 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8_non_immediate(ptr addrspace(1) %out
 ; VI-SDAG-NEXT:    ;;#ASMEND
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v1, s5
-; VI-SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; VI-SDAG-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-SDAG-NEXT:    ;;#ASMSTART
 ; VI-SDAG-NEXT:    v_lshlrev_b64 v[6:7], v[0:1], 1
 ; VI-SDAG-NEXT:    ;;#ASMEND
 ; VI-SDAG-NEXT:    v_mqsad_pk_u16_u8 v[0:1], v[2:3], v4, v[6:7]
+; VI-SDAG-NEXT:    v_mov_b32_e32 v8, s0
+; VI-SDAG-NEXT:    v_mov_b32_e32 v9, s1
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v3, v1
 ; VI-SDAG-NEXT:    v_mov_b32_e32 v2, v0
 ; VI-SDAG-NEXT:    ;;#ASMSTART
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
index b737c318850189..8285348d22cf0f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
@@ -42,13 +42,13 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
 ; GFX803-GISEL-NEXT:    s_mov_b32 s3, 40
 ; GFX803-GISEL-NEXT:    s_mov_b32 s2, 30
 ; GFX803-GISEL-NEXT:    s_mov_b32 s1, 20
-; GFX803-GISEL-NEXT:    s_mov_b32 s0, 10
 ; GFX803-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, s10
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, s11
 ; GFX803-GISEL-NEXT:    ;;#ASMSTART
 ; GFX803-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
+; GFX803-GISEL-NEXT:    s_mov_b32 s0, 10
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v2, s4
@@ -268,13 +268,13 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
 ; GFX803-GISEL-NEXT:    s_mov_b32 s1, 0
 ; GFX803-GISEL-NEXT:    s_mov_b32 s0, 1.0
 ; GFX803-GISEL-NEXT:    s_mov_b32 s2, s1
-; GFX803-GISEL-NEXT:    s_mov_b32 s3, s1
 ; GFX803-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, s10
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, s11
 ; GFX803-GISEL-NEXT:    ;;#ASMSTART
 ; GFX803-GISEL-NEXT:    v_lsrlrev_b64 v[2:3], v[0:1], 1
 ; GFX803-GISEL-NEXT:    ;;#ASMEND
+; GFX803-GISEL-NEXT:    s_mov_b32 s3, s1
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX803-GISEL-NEXT:    v_mov_b32_e32 v2, s4
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
index e56b7dd9f7f6e2..a051e15551ce2a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
@@ -667,8 +667,8 @@ define amdgpu_kernel void @test_readfirstlane_fi(ptr addrspace(1) %out) #0 {
 ; CHECK-SDAG-LABEL: test_readfirstlane_fi:
 ; CHECK-SDAG:       ; %bb.0:
 ; CHECK-SDAG-NEXT:    s_add_u32 s0, s0, s17
-; CHECK-SDAG-NEXT:    s_addc_u32 s1, s1, 0
 ; CHECK-SDAG-NEXT:    s_mov_b32 s4, 0
+; CHECK-SDAG-NEXT:    s_addc_u32 s1, s1, 0
 ; CHECK-SDAG-NEXT:    ;;#ASMSTART
 ; CHECK-SDAG-NEXT:    ; use s4
 ; CHECK-SDAG-NEXT:    ;;#ASMEND
@@ -677,8 +677,8 @@ define amdgpu_kernel void @test_readfirstlane_fi(ptr addrspace(1) %out) #0 {
 ; CHECK-GISEL-LABEL: test_readfirstlane_fi:
 ; CHECK-GISEL:       ; %bb.0:
 ; CHECK-GISEL-NEXT:    s_add_u32 s0, s0, s17
-; CHECK-GISEL-NEXT:    s_addc_u32 s1, s1, 0
 ; CHECK-GISEL-NEXT:    s_mov_b32 s4, 0
+; CHECK-GISEL-NEXT:    s_addc_u32 s1, s1, 0
 ; CHECK-GISEL-NEXT:    ;;#ASMSTART
 ; CHECK-GISEL-NEXT:    ; use s4
 ; CHECK-GISEL-NEXT:    ;;#ASMEND
@@ -1005,6 +1005,13 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
 ; CHECK-SDAG-NEXT:    buffer_load_dword v1, off, s[0:3], s32 offset:4
 ; CHECK-SDAG-NEXT:    buffer_load_dword v27, off, s[0:3], s32
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s64, v30
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s63, v29
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s62, v28
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s60, v26
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s59, v25
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s58, v24
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s57, v23
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s56, v22
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s55, v21
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s54, v20
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s53, v19
@@ -1013,17 +1020,6 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s50, v16
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s49, v15
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s48, v14
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s39, v5
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s38, v4
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s37, v3
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s36, v2
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s63, v29
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s62, v28
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s60, v26
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s59, v25
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s58, v24
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s57, v23
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s56, v22
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s47, v13
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s46, v12
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s45, v11
@@ -1032,6 +1028,10 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s42, v8
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s41, v7
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s40, v6
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s39, v5
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s38, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s37, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s36, v2
 ; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(2)
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s67, v0
 ; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(1)
@@ -1092,15 +1092,6 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s37, v3
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s38, v4
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s39, v5
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s48, v14
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s49, v15
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s50, v16
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s51, v17
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s52, v18
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s53, v19
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s54, v20
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s55, v21
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s64, v30
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s40, v6
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s41, v7
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s42, v8
@@ -1109,6 +1100,14 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s45, v11
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s46, v12
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s47, v13
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s48, v14
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s49, v15
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s50, v16
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s51, v17
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s52, v18
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s53, v19
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s54, v20
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s55, v21
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s56, v22
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s57, v23
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s58, v24
@@ -1117,6 +1116,7 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s61, v27
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s62, v28
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s63, v29
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s64, v30
 ; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(2)
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s65, v0
 ; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(1)
@@ -1450,6 +1450,13 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
 ; CHECK-SDAG-NEXT:    buffer_load_dword v1, off, s[0:3], s32 offset:4
 ; CHECK-SDAG-NEXT:    buffer_load_dword v27, off, s[0:3], s32
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s64, v30
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s63, v29
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s62, v28
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s60, v26
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s59, v25
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s58, v24
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s57, v23
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s56, v22
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s55, v21
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s54, v20
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s53, v19
@@ -1458,17 +1465,6 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s50, v16
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s49, v15
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s48, v14
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s39, v5
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s38, v4
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s37, v3
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s36, v2
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s63, v29
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s62, v28
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s60, v26
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s59, v25
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s58, v24
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s57, v23
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s56, v22
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s47, v13
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s46, v12
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s45, v11
@@ -1477,6 +1473,10 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s42, v8
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s41, v7
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s40, v6
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s39, v5
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s38, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s37, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s36, v2
 ; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(2)
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s67, v0
 ; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(1)
@@ -1537,15 +1537,6 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s37, v3
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s38, v4
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s39, v5
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s48, v14
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s49, v15
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s50, v16
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s51, v17
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s52, v18
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s53, v19
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s54, v20
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s55, v21
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s64, v30
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s40, v6
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s41, v7
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s42, v8
@@ -1554,6 +1545,14 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s45, v11
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s46, v12
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s47, v13
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s48, v14
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s49, v15
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s50, v16
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s51, v17
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s52, v18
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s53, v19
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s54, v20
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s55, v21
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s56, v22
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s57, v23
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s58, v24
@@ -1562,6 +1561,7 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s61, v27
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s62, v28
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s63, v29
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s64, v30
 ; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(2)
 ; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s65, v0
 ; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(1)
diff --git a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll
index 4a06c4e658fe9f..22b58304bd48aa 100644
--- a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll
+++ b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll
@@ -320,13 +320,13 @@ define void @scalar_mov_materializes_frame_index_dead_scc() #0 {
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
 ; GFX12-NEXT:    s_mov_b32 exec_lo, s0
 ; GFX12-NEXT:    v_writelane_b32 v1, s55, 0
-; GFX12-NEXT:    s_add_co_i32 s0, s32, 0x4000
 ; GFX12-NEXT:    v_mov_b32_e32 v0, s32
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    s_mov_b32 s55, s0
+; GFX12-NEXT:    s_add_co_i32 s0, s32, 0x4000
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use alloca0 v0
 ; GFX12-NEXT:    ;;#ASMEND
+; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
+; GFX12-NEXT:    s_mov_b32 s55, s0
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use s55
 ; GFX12-NEXT:    ;;#ASMEND
@@ -346,10 +346,10 @@ define void @scalar_mov_materializes_frame_index_dead_scc() #0 {
 ; GFX8-NEXT:    buffer_store_dword v1, off, s[0:3], s6 ; 4-byte Folded Spill
 ; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX8-NEXT:    v_writelane_b32 v1, s55, 0
-; GFX8-NEXT:    s_lshr_b32 s55, s32, 6
 ; GFX8-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
-; GFX8-NEXT:    s_addk_i32 s55, 0x4040
+; GFX8-NEXT:    s_lshr_b32 s55, s32, 6
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, 64, v0
+; GFX8-NEXT:    s_addk_i32 s55, 0x4040
 ; GFX8-NEXT:    ;;#ASMSTART
 ; GFX8-NEXT:    ; use alloca0 v0
 ; GFX8-NEXT:    ;;#ASMEND
@@ -372,10 +372,10 @@ define void @scalar_mov_materializes_frame_index_dead_scc() #0 {
 ; GFX900-NEXT:    buffer_store_dword v1, off, s[0:3], s6 ; 4-byte Folded Spill
 ; GFX900-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX900-NEXT:    v_writelane_b32 v1, s55, 0
-; GFX900-NEXT:    s_lshr_b32 s55, s32, 6
 ; GFX900-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
-; GFX900-NEXT:    s_addk_i32 s55, 0x4040
+; GFX900-NEXT:    s_lshr_b32 s55, s32, 6
 ; GFX900-NEXT:    v_add_u32_e32 v0, 64, v0
+; GFX900-NEXT:    s_addk_i32 s55, 0x4040
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use alloca0 v0
 ; GFX900-NEXT:    ;;#ASMEND
@@ -1315,12 +1315,12 @@ define void @scalar_mov_materializes_frame_index_available_scc_small_offset_fp()
 ; GFX12-NEXT:    s_addk_co_i32 s32, 0x4040
 ; GFX12-NEXT:    v_writelane_b32 v0, s55, 0
 ; GFX12-NEXT:    s_mov_b32 s55, s33
+; GFX12-NEXT:    s_mov_b32 s32, s33
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use s55
 ; GFX12-NEXT:    ;;#ASMEND
 ; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX12-NEXT:    v_readlane_b32 s55, v0, 0
-; GFX12-NEXT:    s_mov_b32 s32, s33
 ; GFX12-NEXT:    s_xor_saveexec_b32 s1, -1
 ; GFX12-NEXT:    scratch_load_b32 v0, off, s33 offset:16384 ; 4-byte Folded Reload
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
@@ -1507,8 +1507,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
 ; GFX12-NEXT:    s_mov_b32 exec_lo, s0
 ; GFX12-NEXT:    v_writelane_b32 v1, s55, 0
-; GFX12-NEXT:    s_add_co_i32 s55, s32, 0x43ec
 ; GFX12-NEXT:    v_mov_b32_e32 v0, s32
+; GFX12-NEXT:    s_add_co_i32 s55, s32, 0x43ec
 ; GFX12-NEXT:    s_and_b32 s0, 0, exec_lo
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use alloca0 v0
@@ -1533,8 +1533,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
 ; GFX8-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX8-NEXT:    v_writelane_b32 v1, s55, 0
 ; GFX8-NEXT:    s_lshr_b32 s4, s32, 6
-; GFX8-NEXT:    s_add_i32 s55, s4, 0x442c
 ; GFX8-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
+; GFX8-NEXT:    s_add_i32 s55, s4, 0x442c
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, 64, v0
 ; GFX8-NEXT:    ;;#ASMSTART
 ; GFX8-NEXT:    ; use alloca0 v0
@@ -1560,8 +1560,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
 ; GFX900-NEXT:    s_mov_b64 exec, s[4:5]
 ; GFX900-NEXT:    v_writelane_b32 v1, s55, 0
 ; GFX900-NEXT:    s_lshr_b32 s4, s32, 6
-; GFX900-NEXT:    s_add_i32 s55, s4, 0x442c
 ; GFX900-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
+; GFX900-NEXT:    s_add_i32 s55, s4, 0x442c
 ; GFX900-NEXT:    v_add_u32_e32 v0, 64, v0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use alloca0 v0
@@ -1710,14 +1710,14 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
 ; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
 ; GFX12-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX12-NEXT:    v_writelane_b32 v1, s55, 0
-; GFX12-NEXT:    s_add_co_i32 s1, s32, 0x4000
 ; GFX12-NEXT:    v_mov_b32_e32 v0, s32
-; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT:    s_lshl2_add_u32 s55, s0, s1
-; GFX12-NEXT:    s_and_b32 s0, 0, exec_lo
+; GFX12-NEXT:    s_add_co_i32 s1, s32, 0x4000
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use alloca0 v0
 ; GFX12-NEXT:    ;;#ASMEND
+; GFX12-NEXT:    s_wait_alu depctr_sa_sdst(0)
+; GFX12-NEXT:    s_lshl2_add_u32 s55, s0, s1
+; GFX12-NEXT:    s_and_b32 s0, 0, exec_lo
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use s55, scc
 ; GFX12-NEXT:    ;;#ASMEND
@@ -1740,8 +1740,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
 ; GFX8-NEXT:    s_lshl_b32 s4, s16, 2
 ; GFX8-NEXT:    s_lshr_b32 s55, s32, 6
 ; GFX8-NEXT:    s_add_i32 s55, s55, s4
-; GFX8-NEXT:    s_addk_i32 s55, 0x4040
 ; GFX8-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
+; GFX8-NEXT:    s_addk_i32 s55, 0x4040
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, 64, v0
 ; GFX8-NEXT:    ;;#ASMSTART
 ; GFX8-NEXT:    ; use alloca0 v0
@@ -1768,8 +1768,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
 ; GFX900-NEXT:    v_writelane_b32 v1, s55, 0
 ; GFX900-NEXT:    s_lshr_b32 s4, s32, 6
 ; GFX900-NEXT:    s_addk_i32 s4, 0x4040
-; GFX900-NEXT:    s_lshl2_add_u32 s55, s16, s4
 ; GFX900-NEXT:    v_lshrrev_b32_e64 v0, 6, s32
+; GFX900-NEXT:    s_lshl2_add_u32 s55, s16, s4
 ; GFX900-NEXT:    v_add_u32_e32 v0, 64, v0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use alloca0 v0
diff --git a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll
index 6efbfcca484b3e..8221a44d532f6a 100644
--- a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll
+++ b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll
@@ -1625,8 +1625,8 @@ define void @scalar_mov_materializes_frame_index_no_live_scc_no_live_sgprs_gep_i
 ; GFX12-NEXT:    v_writelane_b32 v22, s55, 14
 ; GFX12-NEXT:    v_writelane_b32 v22, s30, 15
 ; GFX12-NEXT:    v_writelane_b32 v22, s31, 16
-; GFX12-NEXT:    s_add_co_i32 s58, s32, 0x4200
 ; GFX12-NEXT:    v_mov_b32_e32 v0, s32
+; GFX12-NEXT:    s_add_co_i32 s58, s32, 0x4200
 ; GFX12-NEXT:    s_and_b32 s0, 0, exec_lo
 ; GFX12-NEXT:    ;;#ASMSTART
 ; GFX12-NEXT:    ; use alloca0 v0
diff --git a/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir
new file mode 100644
index 00000000000000..a740efa0d2c020
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir
@@ -0,0 +1,49 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -amdgpu-sched-strategy=coexec -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Encoded DS reads and reads without MMOs retain the modeled result latency
+# at explicit inline-asm inputs. Generic side-effect/order flags do not change
+# the latency of a register result.
+
+---
+name: encoded_read
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: encoded_read:%bb.0
+  ; CHECK: SU(0): {{.*}}DS_READ_B32_vi_gfx9
+  ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $exec
+    $vgpr2 = DS_READ_B32_vi_gfx9 $vgpr0, 0, 0, implicit $exec :: (load (s32), addrspace 3)
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
+
+---
+name: missing_mmo
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: missing_mmo:%bb.0
+  ; CHECK: SU(0): {{.*}}DS_READ_B32_gfx9
+  ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $exec
+    $vgpr2 = DS_READ_B32_gfx9 $vgpr0, 0, 0, implicit $exec
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
+
+---
+name: atomic_lds_result
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: atomic_lds_result:%bb.0
+  ; CHECK: SU(0): {{.*}}DS_ADD_RTN_U32_gfx9
+  ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $exec
+    $vgpr2 = DS_ADD_RTN_U32_gfx9 $vgpr0, $vgpr1, 0, 0, implicit $exec :: (load store (s32), addrspace 3)
+    INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+    S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir
new file mode 100644
index 00000000000000..92a6879290951b
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir
@@ -0,0 +1,93 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -misched-cluster=false -verify-machineinstrs -verify-misched %s -o - | FileCheck %s
+
+# A real asm input must retain its native LDS producer's modeled latency in
+# default and coexec schedules. Check empty tied and nonempty inputs, including
+# IGLP.
+
+--- |
+  define amdgpu_kernel void @tied_input_default() { ret void }
+  define amdgpu_kernel void @tied_input_coexec() #0 { ret void }
+  define amdgpu_kernel void @explicit_input_coexec() #0 { ret void }
+  define amdgpu_kernel void @iglp_input_coexec() #0 { ret void }
+  attributes #0 = { "amdgpu-sched-strategy"="coexec" }
+...
+
+---
+name: tied_input_default
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: name: tied_input_default
+  ; CHECK: liveins: $vgpr0, $vgpr4, $exec
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: $vgpr2_vgpr3 = DS_READ_B64_TR_B16
+  bb.0:
+    liveins: $vgpr0, $vgpr4, $exec
+
+    $vgpr2_vgpr3 = DS_READ_B64_TR_B16 $vgpr0, 64, 0, implicit $exec :: (load (s64), addrspace 3)
+    INLINEASM &"", attdialect, regdef:VReg_64_Align2, def $vgpr2_vgpr3, reguse tiedto:$0, $vgpr2_vgpr3(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    $vgpr5 = V_ADD_U32_e32 $vgpr2, $vgpr3, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name: tied_input_coexec
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: name: tied_input_coexec
+  ; CHECK: liveins: $vgpr0, $vgpr4, $exec
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: $vgpr2_vgpr3 = DS_READ_B64_TR_B16
+  bb.0:
+    liveins: $vgpr0, $vgpr4, $exec
+
+    $vgpr2_vgpr3 = DS_READ_B64_TR_B16 $vgpr0, 64, 0, implicit $exec :: (load (s64), addrspace 3)
+    INLINEASM &"", attdialect, regdef:VReg_64_Align2, def $vgpr2_vgpr3, reguse tiedto:$0, $vgpr2_vgpr3(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    $vgpr5 = V_ADD_U32_e32 $vgpr2, $vgpr3, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name: explicit_input_coexec
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: name: explicit_input_coexec
+  ; CHECK: liveins: $vgpr0, $vgpr4, $exec
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: $vgpr2 = DS_READ_B32_gfx9
+  bb.0:
+    liveins: $vgpr0, $vgpr4, $exec
+
+    $vgpr2 = DS_READ_B32_gfx9 $vgpr0, 64, 0, implicit $exec :: (load (s32), addrspace 3)
+    INLINEASM &"v_add_u32 $0, $1, 1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    $vgpr5 = V_ADD_U32_e32 1, $vgpr6, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name: iglp_input_coexec
+tracksRegLiveness: true
+body: |
+  ; CHECK-LABEL: name: iglp_input_coexec
+  ; CHECK: $vgpr2_vgpr3 = DS_READ_B64_TR_B16
+  ; CHECK: $vgpr4 = V_ADD_U32_e32 1,
+  bb.0:
+    liveins: $vgpr0, $vgpr4, $exec
+
+    IGLP_OPT 3
+    $vgpr2_vgpr3 = DS_READ_B64_TR_B16 $vgpr0, 64, 0, implicit $exec :: (load (s64), addrspace 3)
+    INLINEASM &"", attdialect, regdef:VReg_64_Align2, def $vgpr2_vgpr3, reguse tiedto:$0, $vgpr2_vgpr3(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+    INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+    $vgpr4 = V_ADD_U32_e32 2, $vgpr4, implicit $exec
+    $vgpr5 = V_ADD_U32_e32 $vgpr2, $vgpr3, implicit $exec
+    S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll b/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll
index c22e48557b8769..ffb0c2f1e6a7ee 100644
--- a/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll
+++ b/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll
@@ -154,42 +154,42 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
 ; GFX906-NEXT:    s_mov_b64 exec, s[34:35]
 ; GFX906-NEXT:    s_waitcnt vmcnt(0)
 ; GFX906-NEXT:    v_readlane_b32 s11, v39, 12
+; GFX906-NEXT:    v_readlane_b32 s12, v39, 13
+; GFX906-NEXT:    v_readlane_b32 s13, v39, 14
+; GFX906-NEXT:    v_readlane_b32 s14, v39, 15
+; GFX906-NEXT:    v_readlane_b32 s15, v39, 16
+; GFX906-NEXT:    v_readlane_b32 s16, v39, 17
+; GFX906-NEXT:    v_readlane_b32 s17, v39, 18
+; GFX906-NEXT:    v_readlane_b32 s18, v39, 19
+; GFX906-NEXT:    v_readlane_b32 s19, v39, 20
+; GFX906-NEXT:    v_readlane_b32 s20, v39, 21
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s11
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s12, v39, 13
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s12
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s13, v39, 14
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s13
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s14, v39, 15
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s14
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s15, v39, 16
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s15
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s16, v39, 17
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s16
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s17, v39, 18
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s17
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s18, v39, 19
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s18
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s19, v39, 20
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s19
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s20, v39, 21
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s20
 ; GFX906-NEXT:    ;;#ASMEND
@@ -257,6 +257,15 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
 ; GFX906-NEXT:    v_readlane_b32 s8, v39, 6
 ; GFX906-NEXT:    v_readlane_b32 s10, v39, 4
 ; GFX906-NEXT:    v_readlane_b32 s16, v39, 22
+; GFX906-NEXT:    v_readlane_b32 s21, v39, 12
+; GFX906-NEXT:    v_readlane_b32 s22, v39, 13
+; GFX906-NEXT:    v_readlane_b32 s23, v39, 14
+; GFX906-NEXT:    v_readlane_b32 s24, v39, 15
+; GFX906-NEXT:    v_readlane_b32 s25, v39, 16
+; GFX906-NEXT:    v_readlane_b32 s26, v39, 17
+; GFX906-NEXT:    v_readlane_b32 s27, v39, 18
+; GFX906-NEXT:    v_readlane_b32 s28, v39, 19
+; GFX906-NEXT:    v_readlane_b32 s29, v39, 20
 ; GFX906-NEXT:    v_readlane_b32 s5, v39, 11
 ; GFX906-NEXT:    v_readlane_b32 s7, v39, 9
 ; GFX906-NEXT:    v_readlane_b32 s9, v39, 7
@@ -267,39 +276,30 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
 ; GFX906-NEXT:    v_readlane_b32 s15, v39, 0
 ; GFX906-NEXT:    v_mov_b32_e32 v31, v40
 ; GFX906-NEXT:    v_readlane_b32 s17, v39, 23
-; GFX906-NEXT:    v_readlane_b32 s21, v39, 12
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s21
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s22, v39, 13
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s22
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s23, v39, 14
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s23
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s24, v39, 15
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s24
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s25, v39, 16
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s25
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s26, v39, 17
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s26
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s27, v39, 18
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s27
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s28, v39, 19
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s28
 ; GFX906-NEXT:    ;;#ASMEND
-; GFX906-NEXT:    v_readlane_b32 s29, v39, 20
 ; GFX906-NEXT:    ;;#ASMSTART
 ; GFX906-NEXT:    ; use s29
 ; GFX906-NEXT:    ;;#ASMEND
@@ -536,42 +536,42 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
 ; GFX908-NEXT:    s_mov_b64 exec, s[34:35]
 ; GFX908-NEXT:    s_waitcnt vmcnt(0)
 ; GFX908-NEXT:    v_readlane_b32 s11, v39, 12
+; GFX908-NEXT:    v_readlane_b32 s12, v39, 13
+; GFX908-NEXT:    v_readlane_b32 s13, v39, 14
+; GFX908-NEXT:    v_readlane_b32 s14, v39, 15
+; GFX908-NEXT:    v_readlane_b32 s15, v39, 16
+; GFX908-NEXT:    v_readlane_b32 s16, v39, 17
+; GFX908-NEXT:    v_readlane_b32 s17, v39, 18
+; GFX908-NEXT:    v_readlane_b32 s18, v39, 19
+; GFX908-NEXT:    v_readlane_b32 s19, v39, 20
+; GFX908-NEXT:    v_readlane_b32 s20, v39, 21
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s11
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s12, v39, 13
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s12
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s13, v39, 14
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s13
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s14, v39, 15
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s14
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s15, v39, 16
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s15
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s16, v39, 17
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s16
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s17, v39, 18
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s17
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s18, v39, 19
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s18
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s19, v39, 20
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s19
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s20, v39, 21
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s20
 ; GFX908-NEXT:    ;;#ASMEND
@@ -639,6 +639,15 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
 ; GFX908-NEXT:    v_readlane_b32 s8, v39, 6
 ; GFX908-NEXT:    v_readlane_b32 s10, v39, 4
 ; GFX908-NEXT:    v_readlane_b32 s16, v39, 22
+; GFX908-NEXT:    v_readlane_b32 s21, v39, 12
+; GFX908-NEXT:    v_readlane_b32 s22, v39, 13
+; GFX908-NEXT:    v_readlane_b32 s23, v39, 14
+; GFX908-NEXT:    v_readlane_b32 s24, v39, 15
+; GFX908-NEXT:    v_readlane_b32 s25, v39, 16
+; GFX908-NEXT:    v_readlane_b32 s26, v39, 17
+; GFX908-NEXT:    v_readlane_b32 s27, v39, 18
+; GFX908-NEXT:    v_readlane_b32 s28, v39, 19
+; GFX908-NEXT:    v_readlane_b32 s29, v39, 20
 ; GFX908-NEXT:    v_readlane_b32 s5, v39, 11
 ; GFX908-NEXT:    v_readlane_b32 s7, v39, 9
 ; GFX908-NEXT:    v_readlane_b32 s9, v39, 7
@@ -649,39 +658,30 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
 ; GFX908-NEXT:    v_readlane_b32 s15, v39, 0
 ; GFX908-NEXT:    v_mov_b32_e32 v31, v40
 ; GFX908-NEXT:    v_readlane_b32 s17, v39, 23
-; GFX908-NEXT:    v_readlane_b32 s21, v39, 12
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s21
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s22, v39, 13
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s22
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s23, v39, 14
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s23
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s24, v39, 15
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s24
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s25, v39, 16
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s25
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s26, v39, 17
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s26
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s27, v39, 18
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s27
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s28, v39, 19
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s28
 ; GFX908-NEXT:    ;;#ASMEND
-; GFX908-NEXT:    v_readlane_b32 s29, v39, 20
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ; use s29
 ; GFX908-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll b/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll
index a85d48a8458c04..302115739d0ea6 100644
--- a/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll
+++ b/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll
@@ -260,18 +260,6 @@ define amdgpu_kernel void @kernel0(ptr addrspace(1) %out, i32 %in) #1 {
 ; CHECK-NEXT:    ;;#ASMEND
 ; CHECK-NEXT:    v_readlane_b32 s0, v22, 38
 ; CHECK-NEXT:    v_readlane_b32 s1, v22, 39
-; CHECK-NEXT:    ;;#ASMSTART
-; CHECK-NEXT:    ; use s[16:31]
-; CHECK-NEXT:    ;;#ASMEND
-; CHECK-NEXT:    ;;#ASMSTART
-; CHECK-NEXT:    ; use vcc
-; CHECK-NEXT:    ;;#ASMEND
-; CHECK-NEXT:    ;;#ASMSTART
-; CHECK-NEXT:    ; use s[40:43]
-; CHECK-NEXT:    ;;#ASMEND
-; CHECK-NEXT:    ;;#ASMSTART
-; CHECK-NEXT:    ; use s[44:51]
-; CHECK-NEXT:    ;;#ASMEND
 ; CHECK-NEXT:    v_readlane_b32 s2, v22, 40
 ; CHECK-NEXT:    v_readlane_b32 s3, v22, 41
 ; CHECK-NEXT:    v_readlane_b32 s4, v22, 42
@@ -287,6 +275,18 @@ define amdgpu_kernel void @kernel0(ptr addrspace(1) %out, i32 %in) #1 {
 ; CHECK-NEXT:    v_readlane_b32 s14, v22, 52
 ; CHECK-NEXT:    v_readlane_b32 s15, v22, 53
 ; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use s[16:31]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use vcc
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use s[40:43]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    ;;#ASMSTART
+; CHECK-NEXT:    ; use s[44:51]
+; CHECK-NEXT:    ;;#ASMEND
+; CHECK-NEXT:    ;;#ASMSTART
 ; CHECK-NEXT:    ; use s[0:15]
 ; CHECK-NEXT:    ;;#ASMEND
 ; CHECK-NEXT:    v_readlane_b32 s0, v22, 54
diff --git a/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll b/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll
index 60ba7d9e699272..4f8ae1b17f2d11 100644
--- a/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll
@@ -14610,9 +14610,9 @@ define void @s_shuffle_v2i64_v8i64__15_2() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s30
 ; GFX900-NEXT:    s_mov_b32 s9, s31
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -14639,9 +14639,9 @@ define void @s_shuffle_v2i64_v8i64__15_2() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s30
 ; GFX90A-NEXT:    s_mov_b32 s9, s31
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -14750,9 +14750,9 @@ define void @s_shuffle_v2i64_v8i64__15_4() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s30
 ; GFX900-NEXT:    s_mov_b32 s9, s31
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -14779,9 +14779,9 @@ define void @s_shuffle_v2i64_v8i64__15_4() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s30
 ; GFX90A-NEXT:    s_mov_b32 s9, s31
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -14810,8 +14810,8 @@ define void @s_shuffle_v2i64_v8i64__15_4() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s30
 ; GFX942-NEXT:    s_mov_b32 s9, s31
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -14845,9 +14845,9 @@ define void @s_shuffle_v2i64_v8i64__15_5() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s12, s30
 ; GFX900-NEXT:    s_mov_b32 s13, s31
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -14874,9 +14874,9 @@ define void @s_shuffle_v2i64_v8i64__15_5() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s12, s30
 ; GFX90A-NEXT:    s_mov_b32 s13, s31
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -15010,9 +15010,9 @@ define void @s_shuffle_v2i64_v8i64__15_6() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s30
 ; GFX942-NEXT:    s_mov_b32 s9, s31
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s10, s12
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -15129,9 +15129,9 @@ define void @s_shuffle_v2i64_v8i64__15_7() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s12, s30
 ; GFX942-NEXT:    s_mov_b32 s13, s31
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -15978,10 +15978,10 @@ define void @s_shuffle_v2i64_v8i64__10_0() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[4:19]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s10, s36
 ; GFX900-NEXT:    s_mov_b32 s11, s37
@@ -16017,10 +16017,10 @@ define void @s_shuffle_v2i64_v8i64__10_0() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[4:19]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s10, s36
 ; GFX90A-NEXT:    s_mov_b32 s11, s37
@@ -16179,8 +16179,8 @@ define void @s_shuffle_v2i64_v8i64__12_0() #0 {
 ; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s10, s16
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s11, s17
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -16708,10 +16708,10 @@ define void @s_shuffle_v2i64_v8i64__10_1() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[4:19]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s10, s38
 ; GFX900-NEXT:    s_mov_b32 s11, s39
@@ -16747,10 +16747,10 @@ define void @s_shuffle_v2i64_v8i64__10_1() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[4:19]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s10, s38
 ; GFX90A-NEXT:    s_mov_b32 s11, s39
@@ -16949,8 +16949,8 @@ define void @s_shuffle_v2i64_v8i64__12_1() #0 {
 ; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s10, s18
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s11, s19
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -17116,16 +17116,16 @@ define void @s_shuffle_v2i64_v8i64__14_1() #0 {
 ; GFX942-NEXT:    s_nop 1
 ; GFX942-NEXT:    v_writelane_b32 v0, s31, 1
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[16:31]
+; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[0:15]
+; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s18
 ; GFX942-NEXT:    s_mov_b32 s15, s19
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -17589,11 +17589,11 @@ define void @s_shuffle_v2i64_v8i64__9_2() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[16:31]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s18
 ; GFX900-NEXT:    s_mov_b32 s9, s19
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -17618,11 +17618,11 @@ define void @s_shuffle_v2i64_v8i64__9_2() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[16:31]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s18
 ; GFX90A-NEXT:    s_mov_b32 s9, s19
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -17673,9 +17673,9 @@ define void @s_shuffle_v2i64_v8i64__10_2() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s10, s20
 ; GFX900-NEXT:    s_mov_b32 s11, s21
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -17700,9 +17700,9 @@ define void @s_shuffle_v2i64_v8i64__10_2() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s10, s20
 ; GFX90A-NEXT:    s_mov_b32 s11, s21
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -17729,9 +17729,9 @@ define void @s_shuffle_v2i64_v8i64__10_2() #0 {
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; def s[4:19]
 ; GFX942-NEXT:    ;;#ASMEND
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s10, s20
 ; GFX942-NEXT:    s_mov_b32 s11, s21
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -17763,11 +17763,11 @@ define void @s_shuffle_v2i64_v8i64__11_2() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[16:31]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s22
 ; GFX900-NEXT:    s_mov_b32 s9, s23
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -17792,11 +17792,11 @@ define void @s_shuffle_v2i64_v8i64__11_2() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[16:31]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s22
 ; GFX90A-NEXT:    s_mov_b32 s9, s23
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -17907,11 +17907,11 @@ define void @s_shuffle_v2i64_v8i64__13_2() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[16:31]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s26
 ; GFX900-NEXT:    s_mov_b32 s9, s27
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -17936,11 +17936,11 @@ define void @s_shuffle_v2i64_v8i64__13_2() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[16:31]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s26
 ; GFX90A-NEXT:    s_mov_b32 s9, s27
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18424,9 +18424,9 @@ define void @s_shuffle_v2i64_v8i64__10_3() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s10, s22
 ; GFX900-NEXT:    s_mov_b32 s11, s23
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -18451,9 +18451,9 @@ define void @s_shuffle_v2i64_v8i64__10_3() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s10, s22
 ; GFX90A-NEXT:    s_mov_b32 s11, s23
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18480,9 +18480,9 @@ define void @s_shuffle_v2i64_v8i64__10_3() #0 {
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; def s[4:19]
 ; GFX942-NEXT:    ;;#ASMEND
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s10, s22
 ; GFX942-NEXT:    s_mov_b32 s11, s23
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -18572,9 +18572,9 @@ define void @s_shuffle_v2i64_v8i64__12_3() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s22
 ; GFX900-NEXT:    s_mov_b32 s15, s23
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -18601,9 +18601,9 @@ define void @s_shuffle_v2i64_v8i64__12_3() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s22
 ; GFX90A-NEXT:    s_mov_b32 s15, s23
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -18710,9 +18710,9 @@ define void @s_shuffle_v2i64_v8i64__14_3() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s18, s22
 ; GFX900-NEXT:    s_mov_b32 s19, s23
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[16:17]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[18:19]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -18739,9 +18739,9 @@ define void @s_shuffle_v2i64_v8i64__14_3() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s18, s22
 ; GFX90A-NEXT:    s_mov_b32 s19, s23
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[16:17]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[18:19]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -19104,11 +19104,11 @@ define void @s_shuffle_v2i64_v8i64__9_4() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[16:31]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s18
 ; GFX900-NEXT:    s_mov_b32 s9, s19
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -19133,11 +19133,11 @@ define void @s_shuffle_v2i64_v8i64__9_4() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[16:31]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s18
 ; GFX90A-NEXT:    s_mov_b32 s9, s19
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -19165,9 +19165,9 @@ define void @s_shuffle_v2i64_v8i64__9_4() #0 {
 ; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s18
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s19
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -19255,11 +19255,11 @@ define void @s_shuffle_v2i64_v8i64__11_4() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[16:31]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s22
 ; GFX900-NEXT:    s_mov_b32 s9, s23
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -19284,11 +19284,11 @@ define void @s_shuffle_v2i64_v8i64__11_4() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[16:31]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s22
 ; GFX90A-NEXT:    s_mov_b32 s9, s23
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -19316,9 +19316,9 @@ define void @s_shuffle_v2i64_v8i64__11_4() #0 {
 ; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s22
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s23
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -19352,9 +19352,9 @@ define void @s_shuffle_v2i64_v8i64__12_4() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s26, s12
 ; GFX900-NEXT:    s_mov_b32 s27, s13
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[24:25]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[26:27]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -19381,9 +19381,9 @@ define void @s_shuffle_v2i64_v8i64__12_4() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s26, s12
 ; GFX90A-NEXT:    s_mov_b32 s27, s13
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[24:25]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[26:27]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -19432,11 +19432,11 @@ define void @s_shuffle_v2i64_v8i64__13_4() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[16:31]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s26
 ; GFX900-NEXT:    s_mov_b32 s9, s27
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -19461,11 +19461,11 @@ define void @s_shuffle_v2i64_v8i64__13_4() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[16:31]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s26
 ; GFX90A-NEXT:    s_mov_b32 s9, s27
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -19493,9 +19493,9 @@ define void @s_shuffle_v2i64_v8i64__13_4() #0 {
 ; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s26
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s27
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -20030,9 +20030,9 @@ define void @s_shuffle_v2i64_v8i64__9_5() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s12, s18
 ; GFX900-NEXT:    s_mov_b32 s13, s19
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -20059,9 +20059,9 @@ define void @s_shuffle_v2i64_v8i64__9_5() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s12, s18
 ; GFX90A-NEXT:    s_mov_b32 s13, s19
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -20168,9 +20168,9 @@ define void @s_shuffle_v2i64_v8i64__11_5() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s12, s22
 ; GFX900-NEXT:    s_mov_b32 s13, s23
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -20197,9 +20197,9 @@ define void @s_shuffle_v2i64_v8i64__11_5() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s12, s22
 ; GFX90A-NEXT:    s_mov_b32 s13, s23
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -20310,9 +20310,9 @@ define void @s_shuffle_v2i64_v8i64__13_5() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s12, s26
 ; GFX900-NEXT:    s_mov_b32 s13, s27
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -20339,9 +20339,9 @@ define void @s_shuffle_v2i64_v8i64__13_5() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s12, s26
 ; GFX90A-NEXT:    s_mov_b32 s13, s27
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -20962,16 +20962,16 @@ define void @s_shuffle_v2i64_v8i64__9_6() #0 {
 ; GFX942-NEXT:    s_nop 1
 ; GFX942-NEXT:    v_writelane_b32 v0, s31, 1
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[16:31]
+; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[0:15]
+; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s18
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s19
 ; GFX942-NEXT:    s_mov_b32 s10, s12
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -21137,16 +21137,16 @@ define void @s_shuffle_v2i64_v8i64__11_6() #0 {
 ; GFX942-NEXT:    s_nop 1
 ; GFX942-NEXT:    v_writelane_b32 v0, s31, 1
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[16:31]
+; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[0:15]
+; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s22
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s23
 ; GFX942-NEXT:    s_mov_b32 s10, s12
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -21362,16 +21362,16 @@ define void @s_shuffle_v2i64_v8i64__13_6() #0 {
 ; GFX942-NEXT:    s_nop 1
 ; GFX942-NEXT:    v_writelane_b32 v0, s31, 1
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[16:31]
+; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[0:15]
+; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s26
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s27
 ; GFX942-NEXT:    s_mov_b32 s10, s12
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -22033,9 +22033,9 @@ define void @s_shuffle_v2i64_v8i64__9_7() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s12, s18
 ; GFX942-NEXT:    s_mov_b32 s13, s19
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -22208,9 +22208,9 @@ define void @s_shuffle_v2i64_v8i64__11_7() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s12, s22
 ; GFX942-NEXT:    s_mov_b32 s13, s23
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -22387,9 +22387,9 @@ define void @s_shuffle_v2i64_v8i64__13_7() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s12, s26
 ; GFX942-NEXT:    s_mov_b32 s13, s27
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -23348,10 +23348,10 @@ define void @s_shuffle_v2i64_v8i64__3_9() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[8:23]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[8:23]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s42
 ; GFX900-NEXT:    s_mov_b32 s9, s43
@@ -23387,10 +23387,10 @@ define void @s_shuffle_v2i64_v8i64__3_9() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[8:23]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[8:23]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s42
 ; GFX90A-NEXT:    s_mov_b32 s9, s43
@@ -23451,9 +23451,9 @@ define void @s_shuffle_v2i64_v8i64__4_9() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s18
 ; GFX900-NEXT:    s_mov_b32 s15, s19
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -23480,9 +23480,9 @@ define void @s_shuffle_v2i64_v8i64__4_9() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s18
 ; GFX90A-NEXT:    s_mov_b32 s15, s19
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -23531,9 +23531,9 @@ define void @s_shuffle_v2i64_v8i64__5_9() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[8:23]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s26
 ; GFX900-NEXT:    s_mov_b32 s9, s27
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -23558,9 +23558,9 @@ define void @s_shuffle_v2i64_v8i64__5_9() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[8:23]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s26
 ; GFX90A-NEXT:    s_mov_b32 s9, s27
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -23587,9 +23587,9 @@ define void @s_shuffle_v2i64_v8i64__5_9() #0 {
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; def s[8:23]
 ; GFX942-NEXT:    ;;#ASMEND
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s8, s26
 ; GFX942-NEXT:    s_mov_b32 s9, s27
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -23706,9 +23706,9 @@ define void @s_shuffle_v2i64_v8i64__6_9() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s18
 ; GFX942-NEXT:    s_mov_b32 s15, s19
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -24159,10 +24159,10 @@ define void @s_shuffle_v2i64_v8i64__1_10() #0 {
 ; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s18
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s19
 ; GFX942-NEXT:    s_mov_b32 s10, s4
 ; GFX942-NEXT:    s_mov_b32 s11, s5
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -24251,10 +24251,10 @@ define void @s_shuffle_v2i64_v8i64__3_10() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[8:23]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[8:23]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s42
 ; GFX900-NEXT:    s_mov_b32 s9, s43
@@ -24292,10 +24292,10 @@ define void @s_shuffle_v2i64_v8i64__3_10() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[8:23]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[8:23]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s42
 ; GFX90A-NEXT:    s_mov_b32 s9, s43
@@ -24360,9 +24360,9 @@ define void @s_shuffle_v2i64_v8i64__4_10() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s20
 ; GFX900-NEXT:    s_mov_b32 s15, s21
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -24389,9 +24389,9 @@ define void @s_shuffle_v2i64_v8i64__4_10() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s20
 ; GFX90A-NEXT:    s_mov_b32 s15, s21
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -24440,11 +24440,11 @@ define void @s_shuffle_v2i64_v8i64__5_10() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[8:23]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s26
 ; GFX900-NEXT:    s_mov_b32 s9, s27
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -24469,11 +24469,11 @@ define void @s_shuffle_v2i64_v8i64__5_10() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[8:23]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s26
 ; GFX90A-NEXT:    s_mov_b32 s9, s27
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -24609,9 +24609,9 @@ define void @s_shuffle_v2i64_v8i64__6_10() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s20
 ; GFX942-NEXT:    s_mov_b32 s15, s21
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -25101,10 +25101,10 @@ define void @s_shuffle_v2i64_v8i64__1_11() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[4:19]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s38
 ; GFX900-NEXT:    s_mov_b32 s9, s39
@@ -25140,10 +25140,10 @@ define void @s_shuffle_v2i64_v8i64__1_11() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[4:19]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s38
 ; GFX90A-NEXT:    s_mov_b32 s9, s39
@@ -25258,9 +25258,9 @@ define void @s_shuffle_v2i64_v8i64__3_11() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s22
 ; GFX900-NEXT:    s_mov_b32 s9, s23
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -25285,9 +25285,9 @@ define void @s_shuffle_v2i64_v8i64__3_11() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s22
 ; GFX90A-NEXT:    s_mov_b32 s9, s23
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -25314,9 +25314,9 @@ define void @s_shuffle_v2i64_v8i64__3_11() #0 {
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; def s[4:19]
 ; GFX942-NEXT:    ;;#ASMEND
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s8, s22
 ; GFX942-NEXT:    s_mov_b32 s9, s23
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -25350,9 +25350,9 @@ define void @s_shuffle_v2i64_v8i64__4_11() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s22
 ; GFX900-NEXT:    s_mov_b32 s15, s23
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -25379,9 +25379,9 @@ define void @s_shuffle_v2i64_v8i64__4_11() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s22
 ; GFX90A-NEXT:    s_mov_b32 s15, s23
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -25571,9 +25571,9 @@ define void @s_shuffle_v2i64_v8i64__6_11() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s22
 ; GFX942-NEXT:    s_mov_b32 s15, s23
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -25987,10 +25987,10 @@ define void @s_shuffle_v2i64_v8i64__1_12() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[4:19]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s38
 ; GFX900-NEXT:    s_mov_b32 s9, s39
@@ -26028,10 +26028,10 @@ define void @s_shuffle_v2i64_v8i64__1_12() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[4:19]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s38
 ; GFX90A-NEXT:    s_mov_b32 s9, s39
@@ -26150,11 +26150,11 @@ define void @s_shuffle_v2i64_v8i64__3_12() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s22
 ; GFX900-NEXT:    s_mov_b32 s9, s23
 ; GFX900-NEXT:    s_mov_b32 s10, s12
 ; GFX900-NEXT:    s_mov_b32 s11, s13
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -26179,11 +26179,11 @@ define void @s_shuffle_v2i64_v8i64__3_12() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s22
 ; GFX90A-NEXT:    s_mov_b32 s9, s23
 ; GFX90A-NEXT:    s_mov_b32 s10, s12
 ; GFX90A-NEXT:    s_mov_b32 s11, s13
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -26210,11 +26210,11 @@ define void @s_shuffle_v2i64_v8i64__3_12() #0 {
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; def s[4:19]
 ; GFX942-NEXT:    ;;#ASMEND
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s8, s22
 ; GFX942-NEXT:    s_mov_b32 s9, s23
 ; GFX942-NEXT:    s_mov_b32 s10, s12
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -26248,9 +26248,9 @@ define void @s_shuffle_v2i64_v8i64__4_12() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s24
 ; GFX900-NEXT:    s_mov_b32 s15, s25
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -26277,9 +26277,9 @@ define void @s_shuffle_v2i64_v8i64__4_12() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s24
 ; GFX90A-NEXT:    s_mov_b32 s15, s25
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -26475,9 +26475,9 @@ define void @s_shuffle_v2i64_v8i64__6_12() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s24
 ; GFX942-NEXT:    s_mov_b32 s15, s25
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -26964,8 +26964,8 @@ define void @s_shuffle_v2i64_v8i64__1_13() #0 {
 ; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s18
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s19
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -27055,9 +27055,9 @@ define void @s_shuffle_v2i64_v8i64__3_13() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s12, s22
 ; GFX900-NEXT:    s_mov_b32 s13, s23
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -27084,9 +27084,9 @@ define void @s_shuffle_v2i64_v8i64__3_13() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s12, s22
 ; GFX90A-NEXT:    s_mov_b32 s13, s23
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -27137,9 +27137,9 @@ define void @s_shuffle_v2i64_v8i64__4_13() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s26
 ; GFX900-NEXT:    s_mov_b32 s15, s27
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -27166,9 +27166,9 @@ define void @s_shuffle_v2i64_v8i64__4_13() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s26
 ; GFX90A-NEXT:    s_mov_b32 s15, s27
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -27362,9 +27362,9 @@ define void @s_shuffle_v2i64_v8i64__6_13() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s26
 ; GFX942-NEXT:    s_mov_b32 s15, s27
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -27846,10 +27846,10 @@ define void @s_shuffle_v2i64_v8i64__1_14() #0 {
 ; GFX900-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX900-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[36:51]
+; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    ;;#ASMSTART
-; GFX900-NEXT:    ; def s[4:19]
+; GFX900-NEXT:    ; def s[36:51]
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s8, s38
 ; GFX900-NEXT:    s_mov_b32 s9, s39
@@ -27887,10 +27887,10 @@ define void @s_shuffle_v2i64_v8i64__1_14() #0 {
 ; GFX90A-NEXT:    v_writelane_b32 v0, s50, 6
 ; GFX90A-NEXT:    v_writelane_b32 v0, s51, 7
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[36:51]
+; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    ;;#ASMSTART
-; GFX90A-NEXT:    ; def s[4:19]
+; GFX90A-NEXT:    ; def s[36:51]
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s8, s38
 ; GFX90A-NEXT:    s_mov_b32 s9, s39
@@ -27929,10 +27929,10 @@ define void @s_shuffle_v2i64_v8i64__1_14() #0 {
 ; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s8, s18
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b32 s9, s19
 ; GFX942-NEXT:    s_mov_b32 s10, s12
 ; GFX942-NEXT:    s_mov_b32 s11, s13
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -28020,11 +28020,11 @@ define void @s_shuffle_v2i64_v8i64__3_14() #0 {
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; def s[4:19]
 ; GFX900-NEXT:    ;;#ASMEND
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b32 s8, s22
 ; GFX900-NEXT:    s_mov_b32 s9, s23
 ; GFX900-NEXT:    s_mov_b32 s10, s16
 ; GFX900-NEXT:    s_mov_b32 s11, s17
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -28049,11 +28049,11 @@ define void @s_shuffle_v2i64_v8i64__3_14() #0 {
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def s[4:19]
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b32 s8, s22
 ; GFX90A-NEXT:    s_mov_b32 s9, s23
 ; GFX90A-NEXT:    s_mov_b32 s10, s16
 ; GFX90A-NEXT:    s_mov_b32 s11, s17
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -28106,9 +28106,9 @@ define void @s_shuffle_v2i64_v8i64__4_14() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s28
 ; GFX900-NEXT:    s_mov_b32 s15, s29
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -28135,9 +28135,9 @@ define void @s_shuffle_v2i64_v8i64__4_14() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s28
 ; GFX90A-NEXT:    s_mov_b32 s15, s29
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -28333,9 +28333,9 @@ define void @s_shuffle_v2i64_v8i64__6_14() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s28
 ; GFX942-NEXT:    s_mov_b32 s15, s29
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -28907,16 +28907,16 @@ define void @s_shuffle_v2i64_v8i64__1_15() #0 {
 ; GFX942-NEXT:    s_nop 1
 ; GFX942-NEXT:    v_writelane_b32 v0, s31, 1
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[16:31]
+; GFX942-NEXT:    ; def s[0:15]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    ;;#ASMSTART
-; GFX942-NEXT:    ; def s[0:15]
+; GFX942-NEXT:    ; def s[16:31]
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s12, s18
 ; GFX942-NEXT:    s_mov_b32 s13, s19
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
@@ -29006,9 +29006,9 @@ define void @s_shuffle_v2i64_v8i64__3_15() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s16, s22
 ; GFX900-NEXT:    s_mov_b32 s17, s23
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[16:17]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[18:19]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -29035,9 +29035,9 @@ define void @s_shuffle_v2i64_v8i64__3_15() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s16, s22
 ; GFX90A-NEXT:    s_mov_b32 s17, s23
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[16:17]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[18:19]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -29090,9 +29090,9 @@ define void @s_shuffle_v2i64_v8i64__4_15() #0 {
 ; GFX900-NEXT:    ;;#ASMEND
 ; GFX900-NEXT:    s_mov_b32 s14, s30
 ; GFX900-NEXT:    s_mov_b32 s15, s31
-; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX900-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX900-NEXT:    ;;#ASMSTART
 ; GFX900-NEXT:    ; use s[8:11]
 ; GFX900-NEXT:    ;;#ASMEND
@@ -29119,9 +29119,9 @@ define void @s_shuffle_v2i64_v8i64__4_15() #0 {
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_mov_b32 s14, s30
 ; GFX90A-NEXT:    s_mov_b32 s15, s31
-; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX90A-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; use s[8:11]
 ; GFX90A-NEXT:    ;;#ASMEND
@@ -29317,9 +29317,9 @@ define void @s_shuffle_v2i64_v8i64__6_15() #0 {
 ; GFX942-NEXT:    ;;#ASMEND
 ; GFX942-NEXT:    s_mov_b32 s14, s30
 ; GFX942-NEXT:    s_mov_b32 s15, s31
-; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    s_mov_b64 s[8:9], s[12:13]
 ; GFX942-NEXT:    s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT:    v_readlane_b32 s30, v0, 0
 ; GFX942-NEXT:    ;;#ASMSTART
 ; GFX942-NEXT:    ; use s[8:11]
 ; GFX942-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/spill-agpr.ll b/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
index 1ca643d90aac02..364ccdde2002ea 100644
--- a/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
+++ b/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
@@ -34,11 +34,11 @@ define amdgpu_kernel void @max_12regs_13a_used(i32 %cond, ptr addrspace(1) %arg,
 ; GFX908-NEXT:    v_accvgpr_read_b32 v2, a5
 ; GFX908-NEXT:    v_accvgpr_read_b32 v3, a6
 ; GFX908-NEXT:    v_accvgpr_read_b32 v4, a7
-; GFX908-NEXT:    v_accvgpr_write_b32 a4, 4
 ; GFX908-NEXT:    v_accvgpr_write_b32 a8, 5
 ; GFX908-NEXT:    v_accvgpr_write_b32 a9, 1
 ; GFX908-NEXT:    v_accvgpr_write_b32 a10, 2
 ; GFX908-NEXT:    v_accvgpr_write_b32 a11, 3
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, 4
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ;;#ASMEND
 ; GFX908-NEXT:    v_accvgpr_write_b32 a7, v4
@@ -77,17 +77,16 @@ define amdgpu_kernel void @max_12regs_13a_used(i32 %cond, ptr addrspace(1) %arg,
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    s_endpgm
 ; GFX90A-NEXT:  .LBB0_2: ; %use
-; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, 5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, 1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, 2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, 3
 ; GFX90A-NEXT:    v_accvgpr_read_b32 v9, a7
 ; GFX90A-NEXT:    v_accvgpr_read_b32 v8, a6
 ; GFX90A-NEXT:    v_accvgpr_read_b32 v7, a5
 ; GFX90A-NEXT:    v_accvgpr_read_b32 v6, a4
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a4, 4
 ; GFX90A-NEXT:    v_mov_b32_e32 v4, 0
-; GFX90A-NEXT:    v_accvgpr_write_b32 a8, 5
-; GFX90A-NEXT:    v_accvgpr_write_b32 a9, 1
-; GFX90A-NEXT:    v_accvgpr_write_b32 a10, 2
-; GFX90A-NEXT:    v_accvgpr_write_b32 a11, 3
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ;;#ASMEND
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a4, v6
@@ -139,13 +138,14 @@ define amdgpu_kernel void @max_10_vgprs_used_9a() #1 {
 ; GFX908-NEXT:    ;;#ASMEND
 ; GFX908-NEXT:    v_accvgpr_read_b32 v5, a2
 ; GFX908-NEXT:    v_accvgpr_read_b32 v4, a1
-; GFX908-NEXT:    v_accvgpr_write_b32 a1, v0
 ; GFX908-NEXT:    v_accvgpr_write_b32 a4, v3
 ; GFX908-NEXT:    v_accvgpr_write_b32 a3, v2
 ; GFX908-NEXT:    v_accvgpr_write_b32 a2, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v0
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ;;#ASMEND
 ; GFX908-NEXT:    v_accvgpr_write_b32 a0, v4
+; GFX908-NEXT:    s_nop 1
 ; GFX908-NEXT:    v_accvgpr_write_b32 a1, v5
 ; GFX908-NEXT:    ;;#ASMSTART
 ; GFX908-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/unspill-vgpr-after-rewrite-vgpr-mfma.ll b/llvm/test/CodeGen/AMDGPU/unspill-vgpr-after-rewrite-vgpr-mfma.ll
index 80f94e87515e49..77af787ec53c5f 100644
--- a/llvm/test/CodeGen/AMDGPU/unspill-vgpr-after-rewrite-vgpr-mfma.ll
+++ b/llvm/test/CodeGen/AMDGPU/unspill-vgpr-after-rewrite-vgpr-mfma.ll
@@ -284,10 +284,10 @@ define void @eliminate_spill_after_mfma_rewrite_x2(i32 %x, i32 %y, <4 x i32> %ar
 ; CHECK-NEXT:    buffer_store_dword v29, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
 ; CHECK-NEXT:    buffer_store_dword v30, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
 ; CHECK-NEXT:    buffer_store_dword v31, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; CHECK-NEXT:    v_accvgpr_read_b32 v2, a2
-; CHECK-NEXT:    v_accvgpr_read_b32 v3, a3
 ; CHECK-NEXT:    v_accvgpr_read_b32 v0, a0
 ; CHECK-NEXT:    v_accvgpr_read_b32 v1, a1
+; CHECK-NEXT:    v_accvgpr_read_b32 v2, a2
+; CHECK-NEXT:    v_accvgpr_read_b32 v3, a3
 ; CHECK-NEXT:    ;;#ASMSTART
 ; CHECK-NEXT:    ; def v[34:37]
 ; CHECK-NEXT:    ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll b/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll
index 2550f4d1fbc85d..4428cd60c9bd02 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll
@@ -65,17 +65,17 @@ define amdgpu_ps <8 x float> @wmma_scale_src_extra_use(<16 x i32> %a, <16 x i32>
 ; CHECK-NEXT:    v_nop
 ; CHECK-NEXT:    global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
 ; CHECK-NEXT:    v_perm_b32 v40, s0, s1, 4
+; CHECK-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; CHECK-NEXT:    v_wmma_scale_f32_16x16x128_f8f6f4 v[32:39], v[0:15], v[16:31], v[32:39], v40, v40 matrix_a_fmt:MATRIX_FMT_BF8 matrix_b_fmt:MATRIX_FMT_BF8
 ; CHECK-NEXT:    ;;#ASMSTART
 ; CHECK-NEXT:    ; use v40
 ; CHECK-NEXT:    ;;#ASMEND
-; CHECK-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(TRANS32_DEP_1)
-; CHECK-NEXT:    v_wmma_scale_f32_16x16x128_f8f6f4 v[32:39], v[0:15], v[16:31], v[32:39], v40, v40 matrix_a_fmt:MATRIX_FMT_BF8 matrix_b_fmt:MATRIX_FMT_BF8
 ; CHECK-NEXT:    v_nop
 ; CHECK-NEXT:    v_nop
 ; CHECK-NEXT:    v_nop
 ; CHECK-NEXT:    v_nop
-; CHECK-NEXT:    v_dual_mov_b32 v0, v32 :: v_dual_mov_b32 v1, v33
 ; CHECK-NEXT:    s_delay_alu instid0(TRANS32_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_1)
+; CHECK-NEXT:    v_dual_mov_b32 v0, v32 :: v_dual_mov_b32 v1, v33
 ; CHECK-NEXT:    v_dual_mov_b32 v2, v34 :: v_dual_mov_b32 v3, v35
 ; CHECK-NEXT:    v_dual_mov_b32 v4, v36 :: v_dual_mov_b32 v5, v37
 ; CHECK-NEXT:    v_dual_mov_b32 v6, v38 :: v_dual_mov_b32 v7, v39



More information about the llvm-commits mailing list