[llvm-branch-commits] [llvm] AMDGPU: Start using AV_MOV_B64_IMM_PSEUDO (PR #154500)
Matt Arsenault via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Wed Aug 20 03:17:18 PDT 2025
https://github.com/arsenm updated https://github.com/llvm/llvm-project/pull/154500
>From 17eb90204f7dee43ab8ce794243e2084ff988f83 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Wed, 20 Aug 2025 18:21:20 +0900
Subject: [PATCH] AMDGPU: Start using AV_MOV_B64_IMM_PSEUDO
---
.../Target/AMDGPU/AMDGPUPrepareAGPRAlloc.cpp | 22 +-
.../AMDGPU/amdgpu-prepare-agpr-alloc.mir | 53 +-
.../AMDGPU/av-split-dead-valno-crash.ll | 52 +-
.../AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll | 16 +-
.../CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll | 578 +++++++++---------
...m.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll | 8 +-
6 files changed, 388 insertions(+), 341 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPrepareAGPRAlloc.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPrepareAGPRAlloc.cpp
index 3b06e9b00ac69..0137b3f5943d7 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUPrepareAGPRAlloc.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPrepareAGPRAlloc.cpp
@@ -34,6 +34,8 @@ class AMDGPUPrepareAGPRAllocImpl {
const SIInstrInfo &TII;
MachineRegisterInfo &MRI;
+ bool isAV64Imm(const MachineOperand &MO) const;
+
public:
AMDGPUPrepareAGPRAllocImpl(const GCNSubtarget &ST, MachineRegisterInfo &MRI)
: TII(*ST.getInstrInfo()), MRI(MRI) {}
@@ -85,11 +87,16 @@ AMDGPUPrepareAGPRAllocPass::run(MachineFunction &MF,
return PreservedAnalyses::all();
}
+bool AMDGPUPrepareAGPRAllocImpl::isAV64Imm(const MachineOperand &MO) const {
+ return MO.isImm() && TII.isLegalAV64PseudoImm(MO.getImm());
+}
+
bool AMDGPUPrepareAGPRAllocImpl::run(MachineFunction &MF) {
if (MRI.isReserved(AMDGPU::AGPR0))
return false;
- const MCInstrDesc &AVImmPseudo = TII.get(AMDGPU::AV_MOV_B32_IMM_PSEUDO);
+ const MCInstrDesc &AVImmPseudo32 = TII.get(AMDGPU::AV_MOV_B32_IMM_PSEUDO);
+ const MCInstrDesc &AVImmPseudo64 = TII.get(AMDGPU::AV_MOV_B64_IMM_PSEUDO);
bool Changed = false;
for (MachineBasicBlock &MBB : MF) {
@@ -98,8 +105,19 @@ bool AMDGPUPrepareAGPRAllocImpl::run(MachineFunction &MF) {
TII.isInlineConstant(MI, 1)) ||
(MI.getOpcode() == AMDGPU::V_ACCVGPR_WRITE_B32_e64 &&
MI.getOperand(1).isImm())) {
- MI.setDesc(AVImmPseudo);
+ MI.setDesc(AVImmPseudo32);
+ Changed = true;
+ continue;
+ }
+
+ // TODO: If only half of the value is rewritable, is it worth splitting it
+ // up?
+ if ((MI.getOpcode() == AMDGPU::V_MOV_B64_e64 ||
+ MI.getOpcode() == AMDGPU::V_MOV_B64_PSEUDO) &&
+ isAV64Imm(MI.getOperand(1))) {
+ MI.setDesc(AVImmPseudo64);
Changed = true;
+ continue;
}
}
}
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-prepare-agpr-alloc.mir b/llvm/test/CodeGen/AMDGPU/amdgpu-prepare-agpr-alloc.mir
index d277c8104fe44..aaacf1d6f793b 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-prepare-agpr-alloc.mir
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-prepare-agpr-alloc.mir
@@ -114,22 +114,22 @@ body: |
; HAS-AGPR-NEXT: liveins: $vgpr0_vgpr1
; HAS-AGPR-NEXT: {{ $}}
; HAS-AGPR-NEXT: [[V_MOV_B64_e64_:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 $vgpr0_vgpr1, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 54, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B64_e64_1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 1, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B64_e64_2:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 64, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B64_e64_3:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 %stack.0, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 65, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B2:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 279172874240, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B3:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 279172874305, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B4:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4290672329938, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B5:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO -9223372036854775808, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B6:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 1042479491, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B7:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4477415320595726336, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B8:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO $vgpr0_vgpr1, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B9:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO %stack.0, implicit $exec
+ ; HAS-AGPR-NEXT: [[AV_MOV_:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 54, implicit $exec
+ ; HAS-AGPR-NEXT: [[AV_MOV_1:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 1, implicit $exec
+ ; HAS-AGPR-NEXT: [[AV_MOV_2:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 64, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B64_e64_1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 %stack.0, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 65, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 279172874240, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B2:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 279172874305, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B3:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4290672329938, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B4:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO -9223372036854775808, implicit $exec
+ ; HAS-AGPR-NEXT: [[AV_MOV_3:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 1042479491, implicit $exec
+ ; HAS-AGPR-NEXT: [[AV_MOV_4:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 4477415320595726336, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B5:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO $vgpr0_vgpr1, implicit $exec
+ ; HAS-AGPR-NEXT: [[V_MOV_B6:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO %stack.0, implicit $exec
; HAS-AGPR-NEXT: {{ $}}
; HAS-AGPR-NEXT: bb.1:
- ; HAS-AGPR-NEXT: [[V_MOV_B64_e64_4:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 3, implicit $exec
+ ; HAS-AGPR-NEXT: [[AV_MOV_5:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 3, implicit $exec
;
; NO-AGPR-LABEL: name: func64
; NO-AGPR: bb.0:
@@ -181,14 +181,23 @@ tracksRegLiveness: true
body: |
bb.0:
liveins: $vgpr0
- ; HAS-AGPR-LABEL: name: func64_no_agprs
- ; HAS-AGPR: liveins: $vgpr0
- ; HAS-AGPR-NEXT: {{ $}}
- ; HAS-AGPR-NEXT: [[V_MOV_B64_e64_:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 1, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4290672329938, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO -9223372036854775808, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B2:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 1042479491, implicit $exec
- ; HAS-AGPR-NEXT: [[V_MOV_B3:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4477415320595726336, implicit $exec
+ ; GFX90A-LABEL: name: func64_no_agprs
+ ; GFX90A: liveins: $vgpr0
+ ; GFX90A-NEXT: {{ $}}
+ ; GFX90A-NEXT: [[V_MOV_B64_e64_:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_e64 1, implicit $exec
+ ; GFX90A-NEXT: [[V_MOV_B:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4290672329938, implicit $exec
+ ; GFX90A-NEXT: [[V_MOV_B1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO -9223372036854775808, implicit $exec
+ ; GFX90A-NEXT: [[V_MOV_B2:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 1042479491, implicit $exec
+ ; GFX90A-NEXT: [[V_MOV_B3:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4477415320595726336, implicit $exec
+ ;
+ ; GFX908-LABEL: name: func64_no_agprs
+ ; GFX908: liveins: $vgpr0
+ ; GFX908-NEXT: {{ $}}
+ ; GFX908-NEXT: [[AV_MOV_:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 1, implicit $exec
+ ; GFX908-NEXT: [[V_MOV_B:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO 4290672329938, implicit $exec
+ ; GFX908-NEXT: [[V_MOV_B1:%[0-9]+]]:vreg_64_align2 = V_MOV_B64_PSEUDO -9223372036854775808, implicit $exec
+ ; GFX908-NEXT: [[AV_MOV_1:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 1042479491, implicit $exec
+ ; GFX908-NEXT: [[AV_MOV_2:%[0-9]+]]:vreg_64_align2 = AV_MOV_B64_IMM_PSEUDO 4477415320595726336, implicit $exec
;
; NO-AGPR-LABEL: name: func64_no_agprs
; NO-AGPR: liveins: $vgpr0
diff --git a/llvm/test/CodeGen/AMDGPU/av-split-dead-valno-crash.ll b/llvm/test/CodeGen/AMDGPU/av-split-dead-valno-crash.ll
index 89fe0ab526a8a..614b1e38a530f 100644
--- a/llvm/test/CodeGen/AMDGPU/av-split-dead-valno-crash.ll
+++ b/llvm/test/CodeGen/AMDGPU/av-split-dead-valno-crash.ll
@@ -16,10 +16,12 @@ define amdgpu_kernel void @vgpr_mfma_pass_av_split_crash(double %arg1, i1 %arg2,
; CHECK-NEXT: s_bitcmp1_b32 s0, 8
; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0
; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[2:3]
-; CHECK-NEXT: s_xor_b64 s[20:21], s[2:3], -1
; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 1, v0
-; CHECK-NEXT: s_and_b64 s[2:3], exec, s[2:3]
; CHECK-NEXT: v_mov_b32_e32 v0, 0x9037ab78
+; CHECK-NEXT: v_accvgpr_write_b32 a3, v1
+; CHECK-NEXT: s_xor_b64 s[20:21], s[2:3], -1
+; CHECK-NEXT: s_and_b64 s[2:3], exec, s[2:3]
+; CHECK-NEXT: v_accvgpr_write_b32 a2, v0
; CHECK-NEXT: v_mov_b32_e32 v3, 0xbe927e4f
; CHECK-NEXT: v_mov_b32_e32 v4, 0x19f4ec90
; CHECK-NEXT: v_mov_b32_e32 v5, 0x3efa01a0
@@ -34,14 +36,14 @@ define amdgpu_kernel void @vgpr_mfma_pass_av_split_crash(double %arg1, i1 %arg2,
; CHECK-NEXT: v_mov_b32_e32 v14, 0x8427b883
; CHECK-NEXT: v_mov_b32_e32 v15, 0x3fae1bb4
; CHECK-NEXT: s_mov_b64 s[22:23], 0
-; CHECK-NEXT: v_mov_b32_e32 v16, 0x57b87036
-; CHECK-NEXT: v_mov_b32_e32 v17, 0x3fb3b136
+; CHECK-NEXT: v_mov_b32_e32 v0, 0x57b87036
+; CHECK-NEXT: v_mov_b32_e32 v1, 0x3fb3b136
; CHECK-NEXT: s_and_b64 s[4:5], exec, s[16:17]
; CHECK-NEXT: v_mov_b32_e32 v18, 0x55555523
; CHECK-NEXT: v_mov_b32_e32 v19, 0xbfd55555
; CHECK-NEXT: s_and_b64 s[6:7], exec, s[18:19]
; CHECK-NEXT: v_mov_b32_e32 v20, 0
-; CHECK-NEXT: ; implicit-def: $agpr0_agpr1
+; CHECK-NEXT: ; implicit-def: $vgpr30_vgpr31
; CHECK-NEXT: ; implicit-def: $vgpr22_vgpr23
; CHECK-NEXT: s_branch .LBB0_2
; CHECK-NEXT: .LBB0_1: ; %Flow9
@@ -61,9 +63,12 @@ define amdgpu_kernel void @vgpr_mfma_pass_av_split_crash(double %arg1, i1 %arg2,
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: v_mov_b64_e32 v[24:25], s[14:15]
; CHECK-NEXT: flat_load_dwordx2 v[24:25], v[24:25]
-; CHECK-NEXT: v_mov_b64_e32 v[26:27], v[0:1]
+; CHECK-NEXT: v_accvgpr_read_b32 v27, a3
+; CHECK-NEXT: v_accvgpr_read_b32 v26, a2
; CHECK-NEXT: v_mov_b64_e32 v[28:29], v[2:3]
-; CHECK-NEXT: v_mov_b64_e32 v[30:31], v[16:17]
+; CHECK-NEXT: v_mov_b64_e32 v[16:17], v[0:1]
+; CHECK-NEXT: v_accvgpr_write_b32 a0, 0
+; CHECK-NEXT: v_accvgpr_write_b32 a1, 0
; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_fmac_f64_e32 v[26:27], 0, v[24:25]
; CHECK-NEXT: v_fmac_f64_e32 v[28:29], 0, v[26:27]
@@ -79,10 +84,9 @@ define amdgpu_kernel void @vgpr_mfma_pass_av_split_crash(double %arg1, i1 %arg2,
; CHECK-NEXT: v_fmac_f64_e32 v[26:27], 0, v[28:29]
; CHECK-NEXT: v_mov_b64_e32 v[28:29], v[14:15]
; CHECK-NEXT: v_fmac_f64_e32 v[28:29], 0, v[26:27]
-; CHECK-NEXT: v_fmac_f64_e32 v[30:31], 0, v[28:29]
+; CHECK-NEXT: v_fmac_f64_e32 v[16:17], 0, v[28:29]
; CHECK-NEXT: v_mov_b64_e32 v[26:27], v[18:19]
-; CHECK-NEXT: v_fmac_f64_e32 v[26:27], 0, v[30:31]
-; CHECK-NEXT: v_mov_b64_e32 v[30:31], 0
+; CHECK-NEXT: v_fmac_f64_e32 v[26:27], 0, v[16:17]
; CHECK-NEXT: s_branch .LBB0_6
; CHECK-NEXT: .LBB0_5: ; %Flow
; CHECK-NEXT: ; in Loop: Header=BB0_6 Depth=2
@@ -91,30 +95,30 @@ define amdgpu_kernel void @vgpr_mfma_pass_av_split_crash(double %arg1, i1 %arg2,
; CHECK-NEXT: .LBB0_6: ; %.preheader1855.i.i.i3329
; CHECK-NEXT: ; Parent Loop BB0_2 Depth=1
; CHECK-NEXT: ; => This Inner Loop Header: Depth=2
-; CHECK-NEXT: v_mov_b64_e32 v[28:29], v[30:31]
+; CHECK-NEXT: v_accvgpr_read_b32 v29, a1
+; CHECK-NEXT: v_accvgpr_read_b32 v28, a0
; CHECK-NEXT: s_mov_b64 s[24:25], -1
; CHECK-NEXT: s_mov_b64 s[8:9], -1
; CHECK-NEXT: s_mov_b64 vcc, s[2:3]
-; CHECK-NEXT: ; implicit-def: $vgpr30_vgpr31
+; CHECK-NEXT: ; implicit-def: $agpr0_agpr1
; CHECK-NEXT: s_cbranch_vccz .LBB0_5
; CHECK-NEXT: ; %bb.7: ; %.lr.ph2070.i.i.i3291
; CHECK-NEXT: ; in Loop: Header=BB0_6 Depth=2
-; CHECK-NEXT: v_accvgpr_read_b32 v31, a1
-; CHECK-NEXT: v_accvgpr_read_b32 v30, a0
+; CHECK-NEXT: v_accvgpr_write_b32 a0, v30
+; CHECK-NEXT: v_accvgpr_write_b32 a1, v31
; CHECK-NEXT: s_mov_b64 s[8:9], s[18:19]
; CHECK-NEXT: s_mov_b64 vcc, s[6:7]
; CHECK-NEXT: s_cbranch_vccz .LBB0_5
; CHECK-NEXT: ; %bb.8: ; %.preheader1856.preheader.i.i.i3325
; CHECK-NEXT: ; in Loop: Header=BB0_6 Depth=2
+; CHECK-NEXT: v_accvgpr_write_b32 a0, v26
; CHECK-NEXT: s_mov_b64 s[24:25], 0
-; CHECK-NEXT: v_mov_b64_e32 v[30:31], v[26:27]
+; CHECK-NEXT: v_accvgpr_write_b32 a1, v27
; CHECK-NEXT: s_mov_b64 s[8:9], 0
; CHECK-NEXT: s_branch .LBB0_5
; CHECK-NEXT: .LBB0_9: ; in Loop: Header=BB0_2 Depth=1
-; CHECK-NEXT: v_mov_b64_e32 v[24:25], s[10:11]
-; CHECK-NEXT: v_accvgpr_write_b32 a0, v24
; CHECK-NEXT: s_mov_b64 s[22:23], 0
-; CHECK-NEXT: v_accvgpr_write_b32 a1, v25
+; CHECK-NEXT: v_mov_b64_e32 v[30:31], s[10:11]
; CHECK-NEXT: s_mov_b64 s[8:9], s[20:21]
; CHECK-NEXT: s_branch .LBB0_15
; CHECK-NEXT: .LBB0_10: ; in Loop: Header=BB0_2 Depth=1
@@ -128,24 +132,22 @@ define amdgpu_kernel void @vgpr_mfma_pass_av_split_crash(double %arg1, i1 %arg2,
; CHECK-NEXT: ; %bb.12: ; %._crit_edge2105.i.i.i2330.loopexit
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: v_cmp_nlg_f64_e64 s[8:9], 0, v[28:29]
-; CHECK-NEXT: v_accvgpr_write_b32 a0, v24
; CHECK-NEXT: v_cndmask_b32_e64 v23, v23, 0, s[16:17]
-; CHECK-NEXT: v_cndmask_b32_e64 v26, 0, 1, s[8:9]
-; CHECK-NEXT: v_mov_b32_e32 v27, v26
-; CHECK-NEXT: s_and_b64 s[8:9], exec, s[16:17]
; CHECK-NEXT: v_cndmask_b32_e64 v22, v22, 0, s[16:17]
-; CHECK-NEXT: global_store_dwordx2 v20, v[26:27], s[12:13]
+; CHECK-NEXT: v_cndmask_b32_e64 v16, 0, 1, s[8:9]
+; CHECK-NEXT: v_mov_b32_e32 v17, v16
+; CHECK-NEXT: s_and_b64 s[8:9], exec, s[16:17]
+; CHECK-NEXT: global_store_dwordx2 v20, v[16:17], s[12:13]
; CHECK-NEXT: s_cselect_b32 s23, s23, 0
; CHECK-NEXT: s_cselect_b32 s22, s22, 0
; CHECK-NEXT: s_mov_b64 s[8:9], -1
; CHECK-NEXT: s_branch .LBB0_14
; CHECK-NEXT: .LBB0_13: ; in Loop: Header=BB0_2 Depth=1
-; CHECK-NEXT: v_accvgpr_write_b32 a0, v24
; CHECK-NEXT: s_mov_b64 s[8:9], 0
; CHECK-NEXT: v_mov_b64_e32 v[22:23], 0
; CHECK-NEXT: .LBB0_14: ; %Flow6
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
-; CHECK-NEXT: v_accvgpr_write_b32 a1, v25
+; CHECK-NEXT: v_mov_b64_e32 v[30:31], v[24:25]
; CHECK-NEXT: .LBB0_15: ; %Flow6
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: s_mov_b64 s[24:25], -1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll
index 8081a15b53bb7..284ced1727b7e 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll
@@ -39,16 +39,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16(<8 x bfloat> %arg0, <8 x
; GCN-NEXT: v_accvgpr_write_b32 a13, s21
; GCN-NEXT: v_accvgpr_write_b32 a14, s22
; GCN-NEXT: v_accvgpr_write_b32 a15, s23
-; GCN-NEXT: v_mov_b64_e32 v[14:15], 0
; GCN-NEXT: v_mov_b32_e32 v16, s16
+; GCN-NEXT: v_mov_b32_e32 v17, s17
; GCN-NEXT: v_mfma_f32_32x32x16_bf16 a[16:31], v[0:3], v[4:7], a[0:15]
+; GCN-NEXT: v_mov_b32_e32 v18, s18
+; GCN-NEXT: v_mov_b32_e32 v19, s19
; GCN-NEXT: v_mov_b32_e32 v0, s20
; GCN-NEXT: v_mov_b32_e32 v1, s21
; GCN-NEXT: v_mov_b32_e32 v2, s22
; GCN-NEXT: v_mov_b32_e32 v3, s23
-; GCN-NEXT: v_mov_b32_e32 v17, s17
-; GCN-NEXT: v_mov_b32_e32 v18, s18
-; GCN-NEXT: v_mov_b32_e32 v19, s19
+; GCN-NEXT: v_mov_b64_e32 v[14:15], 0
; GCN-NEXT: s_nop 4
; GCN-NEXT: global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
; GCN-NEXT: s_waitcnt vmcnt(0)
@@ -112,16 +112,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__flags(<8 x bfloat> %arg0
; GCN-NEXT: v_accvgpr_write_b32 a13, s21
; GCN-NEXT: v_accvgpr_write_b32 a14, s22
; GCN-NEXT: v_accvgpr_write_b32 a15, s23
-; GCN-NEXT: v_mov_b64_e32 v[14:15], 0
; GCN-NEXT: v_mov_b32_e32 v16, s16
+; GCN-NEXT: v_mov_b32_e32 v17, s17
; GCN-NEXT: v_mfma_f32_32x32x16_bf16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
+; GCN-NEXT: v_mov_b32_e32 v18, s18
+; GCN-NEXT: v_mov_b32_e32 v19, s19
; GCN-NEXT: v_mov_b32_e32 v0, s20
; GCN-NEXT: v_mov_b32_e32 v1, s21
; GCN-NEXT: v_mov_b32_e32 v2, s22
; GCN-NEXT: v_mov_b32_e32 v3, s23
-; GCN-NEXT: v_mov_b32_e32 v17, s17
-; GCN-NEXT: v_mov_b32_e32 v18, s18
-; GCN-NEXT: v_mov_b32_e32 v19, s19
+; GCN-NEXT: v_mov_b64_e32 v[14:15], 0
; GCN-NEXT: s_nop 4
; GCN-NEXT: global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
; GCN-NEXT: s_waitcnt vmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll
index d81ec1c61634b..078a043b94604 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll
@@ -406,16 +406,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
; SDAG-NEXT: v_accvgpr_write_b32 a13, s21
; SDAG-NEXT: v_accvgpr_write_b32 a14, s22
; SDAG-NEXT: v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT: v_mov_b64_e32 v[14:15], 0
; SDAG-NEXT: v_mov_b32_e32 v16, s16
+; SDAG-NEXT: v_mov_b32_e32 v17, s17
; SDAG-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15]
+; SDAG-NEXT: v_mov_b32_e32 v18, s18
+; SDAG-NEXT: v_mov_b32_e32 v19, s19
; SDAG-NEXT: v_mov_b32_e32 v0, s20
; SDAG-NEXT: v_mov_b32_e32 v1, s21
; SDAG-NEXT: v_mov_b32_e32 v2, s22
; SDAG-NEXT: v_mov_b32_e32 v3, s23
-; SDAG-NEXT: v_mov_b32_e32 v17, s17
-; SDAG-NEXT: v_mov_b32_e32 v18, s18
-; SDAG-NEXT: v_mov_b32_e32 v19, s19
+; SDAG-NEXT: v_mov_b64_e32 v[14:15], 0
; SDAG-NEXT: s_nop 4
; SDAG-NEXT: global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
@@ -449,9 +449,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
; GISEL: ; %bb.0:
; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24
; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
-; GISEL-NEXT: v_mov_b64_e32 v[20:21], 0
-; GISEL-NEXT: v_mov_b64_e32 v[26:27], 48
-; GISEL-NEXT: v_mov_b64_e32 v[22:23], 16
+; GISEL-NEXT: v_mov_b64_e32 v[12:13], 0
+; GISEL-NEXT: v_mov_b64_e32 v[14:15], 16
+; GISEL-NEXT: v_mov_b64_e32 v[16:17], 32
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25]
; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27]
@@ -474,31 +474,34 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
; GISEL-NEXT: v_accvgpr_write_b32 a14, s22
; GISEL-NEXT: v_accvgpr_write_b32 a15, s23
; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9]
-; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[18:19], 48
; GISEL-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15]
-; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
-; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17]
-; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
-; GISEL-NEXT: v_mov_b64_e32 v[24:25], 32
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[14:15]
; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11]
-; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15]
-; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19]
-; GISEL-NEXT: s_nop 4
-; GISEL-NEXT: global_store_dwordx4 v[20:21], a[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 7
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: global_store_dwordx4 v[12:13], a[16:19], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], a[20:23], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], a[20:23], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], a[24:27], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[16:17], a[24:27], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], a[28:31], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[18:19], a[28:31], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[20:21], v[8:11], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[12:13], v[8:11], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], v[12:15], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], v[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[16:17]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[18:19]
+; GISEL-NEXT: global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], v[0:3], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
+; GISEL-NEXT: global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_endpgm
;
@@ -530,16 +533,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21
; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22
; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23
-; HEURRC-NEXT: v_mov_b64_e32 v[14:15], 0
; HEURRC-NEXT: v_mov_b32_e32 v16, s16
+; HEURRC-NEXT: v_mov_b32_e32 v17, s17
; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15]
+; HEURRC-NEXT: v_mov_b32_e32 v18, s18
+; HEURRC-NEXT: v_mov_b32_e32 v19, s19
; HEURRC-NEXT: v_mov_b32_e32 v0, s20
; HEURRC-NEXT: v_mov_b32_e32 v1, s21
; HEURRC-NEXT: v_mov_b32_e32 v2, s22
; HEURRC-NEXT: v_mov_b32_e32 v3, s23
-; HEURRC-NEXT: v_mov_b32_e32 v17, s17
-; HEURRC-NEXT: v_mov_b32_e32 v18, s18
-; HEURRC-NEXT: v_mov_b32_e32 v19, s19
+; HEURRC-NEXT: v_mov_b64_e32 v[14:15], 0
; HEURRC-NEXT: s_nop 4
; HEURRC-NEXT: global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
@@ -589,12 +592,12 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
; VGPRRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19]
; VGPRRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21]
; VGPRRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT: v_mov_b64_e32 v[46:47], 0
; VGPRRC-NEXT: v_mov_b32_e32 v48, s16
-; VGPRRC-NEXT: v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15]
; VGPRRC-NEXT: v_mov_b32_e32 v49, s17
+; VGPRRC-NEXT: v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15]
; VGPRRC-NEXT: v_mov_b32_e32 v50, s18
; VGPRRC-NEXT: v_mov_b32_e32 v51, s19
+; VGPRRC-NEXT: v_mov_b64_e32 v[46:47], 0
; VGPRRC-NEXT: s_nop 7
; VGPRRC-NEXT: s_nop 0
; VGPRRC-NEXT: global_store_dwordx4 v[40:41], v[28:31], off sc0 sc1
@@ -605,12 +608,12 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: global_store_dwordx4 v[46:47], v[16:19], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[42:43], v[48:51], off sc0 sc1
-; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: v_mov_b32_e32 v0, s20
; VGPRRC-NEXT: v_mov_b32_e32 v1, s21
; VGPRRC-NEXT: v_mov_b32_e32 v2, s22
; VGPRRC-NEXT: v_mov_b32_e32 v3, s23
+; VGPRRC-NEXT: global_store_dwordx4 v[42:43], v[48:51], off sc0 sc1
+; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: global_store_dwordx4 v[40:41], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: s_nop 0
@@ -788,16 +791,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
; SDAG-NEXT: v_accvgpr_write_b32 a13, s21
; SDAG-NEXT: v_accvgpr_write_b32 a14, s22
; SDAG-NEXT: v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT: v_mov_b64_e32 v[14:15], 0
; SDAG-NEXT: v_mov_b32_e32 v16, s16
+; SDAG-NEXT: v_mov_b32_e32 v17, s17
; SDAG-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
+; SDAG-NEXT: v_mov_b32_e32 v18, s18
+; SDAG-NEXT: v_mov_b32_e32 v19, s19
; SDAG-NEXT: v_mov_b32_e32 v0, s20
; SDAG-NEXT: v_mov_b32_e32 v1, s21
; SDAG-NEXT: v_mov_b32_e32 v2, s22
; SDAG-NEXT: v_mov_b32_e32 v3, s23
-; SDAG-NEXT: v_mov_b32_e32 v17, s17
-; SDAG-NEXT: v_mov_b32_e32 v18, s18
-; SDAG-NEXT: v_mov_b32_e32 v19, s19
+; SDAG-NEXT: v_mov_b64_e32 v[14:15], 0
; SDAG-NEXT: s_nop 4
; SDAG-NEXT: global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
@@ -831,9 +834,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
; GISEL: ; %bb.0:
; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24
; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
-; GISEL-NEXT: v_mov_b64_e32 v[20:21], 0
-; GISEL-NEXT: v_mov_b64_e32 v[26:27], 48
-; GISEL-NEXT: v_mov_b64_e32 v[22:23], 16
+; GISEL-NEXT: v_mov_b64_e32 v[12:13], 0
+; GISEL-NEXT: v_mov_b64_e32 v[14:15], 16
+; GISEL-NEXT: v_mov_b64_e32 v[16:17], 32
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25]
; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27]
@@ -856,31 +859,34 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
; GISEL-NEXT: v_accvgpr_write_b32 a14, s22
; GISEL-NEXT: v_accvgpr_write_b32 a15, s23
; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9]
-; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[18:19], 48
; GISEL-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
-; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
-; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17]
-; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
-; GISEL-NEXT: v_mov_b64_e32 v[24:25], 32
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[14:15]
; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11]
-; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15]
-; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19]
-; GISEL-NEXT: s_nop 4
-; GISEL-NEXT: global_store_dwordx4 v[20:21], a[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 7
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: global_store_dwordx4 v[12:13], a[16:19], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], a[20:23], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], a[20:23], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], a[24:27], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[16:17], a[24:27], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], a[28:31], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[18:19], a[28:31], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[20:21], v[8:11], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[12:13], v[8:11], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], v[12:15], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], v[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[16:17]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[18:19]
+; GISEL-NEXT: global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], v[0:3], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
+; GISEL-NEXT: global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_endpgm
;
@@ -912,16 +918,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21
; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22
; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23
-; HEURRC-NEXT: v_mov_b64_e32 v[14:15], 0
; HEURRC-NEXT: v_mov_b32_e32 v16, s16
+; HEURRC-NEXT: v_mov_b32_e32 v17, s17
; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
+; HEURRC-NEXT: v_mov_b32_e32 v18, s18
+; HEURRC-NEXT: v_mov_b32_e32 v19, s19
; HEURRC-NEXT: v_mov_b32_e32 v0, s20
; HEURRC-NEXT: v_mov_b32_e32 v1, s21
; HEURRC-NEXT: v_mov_b32_e32 v2, s22
; HEURRC-NEXT: v_mov_b32_e32 v3, s23
-; HEURRC-NEXT: v_mov_b32_e32 v17, s17
-; HEURRC-NEXT: v_mov_b32_e32 v18, s18
-; HEURRC-NEXT: v_mov_b32_e32 v19, s19
+; HEURRC-NEXT: v_mov_b64_e32 v[14:15], 0
; HEURRC-NEXT: s_nop 4
; HEURRC-NEXT: global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
@@ -971,12 +977,12 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
; VGPRRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19]
; VGPRRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21]
; VGPRRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT: v_mov_b64_e32 v[46:47], 0
; VGPRRC-NEXT: v_mov_b32_e32 v48, s16
-; VGPRRC-NEXT: v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15] cbsz:2 abid:3 blgp:1
; VGPRRC-NEXT: v_mov_b32_e32 v49, s17
+; VGPRRC-NEXT: v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15] cbsz:2 abid:3 blgp:1
; VGPRRC-NEXT: v_mov_b32_e32 v50, s18
; VGPRRC-NEXT: v_mov_b32_e32 v51, s19
+; VGPRRC-NEXT: v_mov_b64_e32 v[46:47], 0
; VGPRRC-NEXT: s_nop 7
; VGPRRC-NEXT: s_nop 0
; VGPRRC-NEXT: global_store_dwordx4 v[40:41], v[28:31], off sc0 sc1
@@ -987,12 +993,12 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: global_store_dwordx4 v[46:47], v[16:19], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[42:43], v[48:51], off sc0 sc1
-; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: v_mov_b32_e32 v0, s20
; VGPRRC-NEXT: v_mov_b32_e32 v1, s21
; VGPRRC-NEXT: v_mov_b32_e32 v2, s22
; VGPRRC-NEXT: v_mov_b32_e32 v3, s23
+; VGPRRC-NEXT: global_store_dwordx4 v[42:43], v[48:51], off sc0 sc1
+; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: global_store_dwordx4 v[40:41], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: s_nop 0
@@ -2978,47 +2984,47 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
; SDAG-NEXT: v_mov_b64_e32 v[0:1], 48
; SDAG-NEXT: v_mov_b64_e32 v[2:3], 32
-; SDAG-NEXT: v_mov_b64_e32 v[4:5], 16
; SDAG-NEXT: s_waitcnt lgkmcnt(0)
-; SDAG-NEXT: v_mov_b32_e32 v8, s24
-; SDAG-NEXT: v_mov_b32_e32 v9, s25
-; SDAG-NEXT: v_mov_b32_e32 v10, s26
-; SDAG-NEXT: v_mov_b32_e32 v11, s27
-; SDAG-NEXT: v_accvgpr_write_b32 a0, s8
-; SDAG-NEXT: v_mov_b32_e32 v12, s28
-; SDAG-NEXT: v_mov_b32_e32 v13, s29
-; SDAG-NEXT: v_mov_b32_e32 v14, s30
-; SDAG-NEXT: v_mov_b32_e32 v15, s31
-; SDAG-NEXT: v_accvgpr_write_b32 a1, s9
-; SDAG-NEXT: v_accvgpr_write_b32 a2, s10
-; SDAG-NEXT: v_accvgpr_write_b32 a3, s11
-; SDAG-NEXT: v_accvgpr_write_b32 a4, s12
-; SDAG-NEXT: v_accvgpr_write_b32 a5, s13
-; SDAG-NEXT: v_accvgpr_write_b32 a6, s14
-; SDAG-NEXT: v_accvgpr_write_b32 a7, s15
-; SDAG-NEXT: v_accvgpr_write_b32 a8, s16
-; SDAG-NEXT: v_accvgpr_write_b32 a9, s17
-; SDAG-NEXT: v_accvgpr_write_b32 a10, s18
-; SDAG-NEXT: v_accvgpr_write_b32 a11, s19
-; SDAG-NEXT: v_accvgpr_write_b32 a12, s20
-; SDAG-NEXT: v_accvgpr_write_b32 a13, s21
-; SDAG-NEXT: v_accvgpr_write_b32 a14, s22
-; SDAG-NEXT: v_accvgpr_write_b32 a15, s23
+; SDAG-NEXT: v_mov_b32_e32 v4, s24
+; SDAG-NEXT: v_mov_b32_e32 v5, s25
+; SDAG-NEXT: v_mov_b32_e32 v6, s26
+; SDAG-NEXT: v_mov_b32_e32 v7, s27
+; SDAG-NEXT: v_accvgpr_write_b32 a31, s23
+; SDAG-NEXT: v_mov_b32_e32 v8, s28
+; SDAG-NEXT: v_mov_b32_e32 v9, s29
+; SDAG-NEXT: v_mov_b32_e32 v10, s30
+; SDAG-NEXT: v_mov_b32_e32 v11, s31
+; SDAG-NEXT: v_accvgpr_write_b32 a30, s22
+; SDAG-NEXT: v_accvgpr_write_b32 a29, s21
+; SDAG-NEXT: v_accvgpr_write_b32 a28, s20
+; SDAG-NEXT: v_accvgpr_write_b32 a27, s19
+; SDAG-NEXT: v_accvgpr_write_b32 a26, s18
+; SDAG-NEXT: v_accvgpr_write_b32 a25, s17
+; SDAG-NEXT: v_accvgpr_write_b32 a24, s16
+; SDAG-NEXT: v_accvgpr_write_b32 a23, s15
+; SDAG-NEXT: v_accvgpr_write_b32 a22, s14
+; SDAG-NEXT: v_accvgpr_write_b32 a21, s13
+; SDAG-NEXT: v_accvgpr_write_b32 a20, s12
+; SDAG-NEXT: v_accvgpr_write_b32 a19, s11
+; SDAG-NEXT: v_accvgpr_write_b32 a18, s10
+; SDAG-NEXT: v_accvgpr_write_b32 a17, s9
+; SDAG-NEXT: v_accvgpr_write_b32 a16, s8
+; SDAG-NEXT: s_nop 1
+; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[4:7], v[8:11], a[16:31]
+; SDAG-NEXT: v_mov_b64_e32 v[4:5], 16
; SDAG-NEXT: v_mov_b64_e32 v[6:7], 0
-; SDAG-NEXT: s_nop 0
-; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15]
; SDAG-NEXT: v_mov_b32_e32 v8, s16
; SDAG-NEXT: v_mov_b32_e32 v9, s17
; SDAG-NEXT: v_mov_b32_e32 v10, s18
; SDAG-NEXT: v_mov_b32_e32 v11, s19
-; SDAG-NEXT: s_nop 7
-; SDAG-NEXT: global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
+; SDAG-NEXT: s_nop 5
+; SDAG-NEXT: global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
-; SDAG-NEXT: global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
+; SDAG-NEXT: global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
-; SDAG-NEXT: global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
+; SDAG-NEXT: global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
-; SDAG-NEXT: global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
+; SDAG-NEXT: global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
; SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
@@ -3047,9 +3053,9 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
; GISEL: ; %bb.0:
; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24
; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
-; GISEL-NEXT: v_mov_b64_e32 v[20:21], 0
-; GISEL-NEXT: v_mov_b64_e32 v[26:27], 48
-; GISEL-NEXT: v_mov_b64_e32 v[22:23], 16
+; GISEL-NEXT: v_mov_b64_e32 v[12:13], 0
+; GISEL-NEXT: v_mov_b64_e32 v[14:15], 16
+; GISEL-NEXT: v_mov_b64_e32 v[16:17], 32
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25]
; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27]
@@ -3072,31 +3078,34 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
; GISEL-NEXT: v_accvgpr_write_b32 a14, s22
; GISEL-NEXT: v_accvgpr_write_b32 a15, s23
; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9]
-; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[18:19], 48
; GISEL-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15]
-; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
-; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17]
-; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
-; GISEL-NEXT: v_mov_b64_e32 v[24:25], 32
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[14:15]
; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11]
-; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15]
-; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19]
-; GISEL-NEXT: s_nop 4
-; GISEL-NEXT: global_store_dwordx4 v[20:21], a[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 7
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: global_store_dwordx4 v[12:13], a[16:19], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], a[20:23], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], a[20:23], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], a[24:27], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[16:17], a[24:27], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], a[28:31], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[18:19], a[28:31], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[20:21], v[8:11], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[12:13], v[8:11], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], v[12:15], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], v[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[16:17]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[18:19]
+; GISEL-NEXT: global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], v[0:3], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
+; GISEL-NEXT: global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_endpgm
;
@@ -3106,47 +3115,47 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
; HEURRC-NEXT: v_mov_b64_e32 v[0:1], 48
; HEURRC-NEXT: v_mov_b64_e32 v[2:3], 32
-; HEURRC-NEXT: v_mov_b64_e32 v[4:5], 16
; HEURRC-NEXT: s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT: v_mov_b32_e32 v8, s24
-; HEURRC-NEXT: v_mov_b32_e32 v9, s25
-; HEURRC-NEXT: v_mov_b32_e32 v10, s26
-; HEURRC-NEXT: v_mov_b32_e32 v11, s27
-; HEURRC-NEXT: v_accvgpr_write_b32 a0, s8
-; HEURRC-NEXT: v_mov_b32_e32 v12, s28
-; HEURRC-NEXT: v_mov_b32_e32 v13, s29
-; HEURRC-NEXT: v_mov_b32_e32 v14, s30
-; HEURRC-NEXT: v_mov_b32_e32 v15, s31
-; HEURRC-NEXT: v_accvgpr_write_b32 a1, s9
-; HEURRC-NEXT: v_accvgpr_write_b32 a2, s10
-; HEURRC-NEXT: v_accvgpr_write_b32 a3, s11
-; HEURRC-NEXT: v_accvgpr_write_b32 a4, s12
-; HEURRC-NEXT: v_accvgpr_write_b32 a5, s13
-; HEURRC-NEXT: v_accvgpr_write_b32 a6, s14
-; HEURRC-NEXT: v_accvgpr_write_b32 a7, s15
-; HEURRC-NEXT: v_accvgpr_write_b32 a8, s16
-; HEURRC-NEXT: v_accvgpr_write_b32 a9, s17
-; HEURRC-NEXT: v_accvgpr_write_b32 a10, s18
-; HEURRC-NEXT: v_accvgpr_write_b32 a11, s19
-; HEURRC-NEXT: v_accvgpr_write_b32 a12, s20
-; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21
-; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22
-; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23
+; HEURRC-NEXT: v_mov_b32_e32 v4, s24
+; HEURRC-NEXT: v_mov_b32_e32 v5, s25
+; HEURRC-NEXT: v_mov_b32_e32 v6, s26
+; HEURRC-NEXT: v_mov_b32_e32 v7, s27
+; HEURRC-NEXT: v_accvgpr_write_b32 a31, s23
+; HEURRC-NEXT: v_mov_b32_e32 v8, s28
+; HEURRC-NEXT: v_mov_b32_e32 v9, s29
+; HEURRC-NEXT: v_mov_b32_e32 v10, s30
+; HEURRC-NEXT: v_mov_b32_e32 v11, s31
+; HEURRC-NEXT: v_accvgpr_write_b32 a30, s22
+; HEURRC-NEXT: v_accvgpr_write_b32 a29, s21
+; HEURRC-NEXT: v_accvgpr_write_b32 a28, s20
+; HEURRC-NEXT: v_accvgpr_write_b32 a27, s19
+; HEURRC-NEXT: v_accvgpr_write_b32 a26, s18
+; HEURRC-NEXT: v_accvgpr_write_b32 a25, s17
+; HEURRC-NEXT: v_accvgpr_write_b32 a24, s16
+; HEURRC-NEXT: v_accvgpr_write_b32 a23, s15
+; HEURRC-NEXT: v_accvgpr_write_b32 a22, s14
+; HEURRC-NEXT: v_accvgpr_write_b32 a21, s13
+; HEURRC-NEXT: v_accvgpr_write_b32 a20, s12
+; HEURRC-NEXT: v_accvgpr_write_b32 a19, s11
+; HEURRC-NEXT: v_accvgpr_write_b32 a18, s10
+; HEURRC-NEXT: v_accvgpr_write_b32 a17, s9
+; HEURRC-NEXT: v_accvgpr_write_b32 a16, s8
+; HEURRC-NEXT: s_nop 1
+; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[4:7], v[8:11], a[16:31]
+; HEURRC-NEXT: v_mov_b64_e32 v[4:5], 16
; HEURRC-NEXT: v_mov_b64_e32 v[6:7], 0
-; HEURRC-NEXT: s_nop 0
-; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15]
; HEURRC-NEXT: v_mov_b32_e32 v8, s16
; HEURRC-NEXT: v_mov_b32_e32 v9, s17
; HEURRC-NEXT: v_mov_b32_e32 v10, s18
; HEURRC-NEXT: v_mov_b32_e32 v11, s19
-; HEURRC-NEXT: s_nop 7
-; HEURRC-NEXT: global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
+; HEURRC-NEXT: s_nop 5
+; HEURRC-NEXT: global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
-; HEURRC-NEXT: global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
+; HEURRC-NEXT: global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
-; HEURRC-NEXT: global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
+; HEURRC-NEXT: global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
-; HEURRC-NEXT: global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
+; HEURRC-NEXT: global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
; HEURRC-NEXT: global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
@@ -3177,37 +3186,40 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
; VGPRRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
; VGPRRC-NEXT: v_mov_b64_e32 v[32:33], 48
; VGPRRC-NEXT: v_mov_b64_e32 v[34:35], 32
-; VGPRRC-NEXT: v_mov_b64_e32 v[36:37], 16
; VGPRRC-NEXT: s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT: v_mov_b32_e32 v40, s24
-; VGPRRC-NEXT: v_mov_b32_e32 v41, s25
-; VGPRRC-NEXT: v_mov_b32_e32 v42, s26
-; VGPRRC-NEXT: v_mov_b32_e32 v43, s27
-; VGPRRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9]
-; VGPRRC-NEXT: v_mov_b32_e32 v44, s28
-; VGPRRC-NEXT: v_mov_b32_e32 v45, s29
-; VGPRRC-NEXT: v_mov_b32_e32 v46, s30
-; VGPRRC-NEXT: v_mov_b32_e32 v47, s31
-; VGPRRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11]
-; VGPRRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13]
-; VGPRRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15]
-; VGPRRC-NEXT: v_mov_b64_e32 v[8:9], s[16:17]
-; VGPRRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19]
-; VGPRRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21]
-; VGPRRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT: v_mov_b64_e32 v[38:39], 0
-; VGPRRC-NEXT: s_nop 0
-; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[16:31], v[40:43], v[44:47], v[0:15]
+; VGPRRC-NEXT: v_mov_b32_e32 v36, s24
+; VGPRRC-NEXT: v_mov_b32_e32 v37, s25
+; VGPRRC-NEXT: v_mov_b32_e32 v38, s26
+; VGPRRC-NEXT: v_mov_b32_e32 v39, s27
+; VGPRRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23]
+; VGPRRC-NEXT: v_mov_b32_e32 v40, s28
+; VGPRRC-NEXT: v_mov_b32_e32 v41, s29
+; VGPRRC-NEXT: v_mov_b32_e32 v42, s30
+; VGPRRC-NEXT: v_mov_b32_e32 v43, s31
+; VGPRRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21]
+; VGPRRC-NEXT: v_mov_b64_e32 v[26:27], s[18:19]
+; VGPRRC-NEXT: v_mov_b64_e32 v[24:25], s[16:17]
+; VGPRRC-NEXT: v_mov_b64_e32 v[22:23], s[14:15]
+; VGPRRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13]
+; VGPRRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11]
+; VGPRRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9]
+; VGPRRC-NEXT: s_nop 1
+; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[36:39], v[40:43], v[16:31]
; VGPRRC-NEXT: s_nop 7
; VGPRRC-NEXT: s_nop 3
-; VGPRRC-NEXT: global_store_dwordx4 v[32:33], v[28:31], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[32:33], v[12:15], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[34:35], v[24:27], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[34:35], v[8:11], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[36:37], v[20:23], off sc0 sc1
+; VGPRRC-NEXT: s_nop 0
+; VGPRRC-NEXT: v_mov_b64_e32 v[8:9], 16
+; VGPRRC-NEXT: global_store_dwordx4 v[8:9], v[4:7], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[38:39], v[16:19], off sc0 sc1
+; VGPRRC-NEXT: s_nop 0
+; VGPRRC-NEXT: v_mov_b64_e32 v[4:5], 0
+; VGPRRC-NEXT: global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
+; VGPRRC-NEXT: s_nop 0
; VGPRRC-NEXT: v_mov_b32_e32 v0, s16
; VGPRRC-NEXT: v_mov_b32_e32 v1, s17
; VGPRRC-NEXT: v_mov_b32_e32 v2, s18
@@ -3226,14 +3238,14 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
; VGPRRC-NEXT: v_mov_b32_e32 v1, s9
; VGPRRC-NEXT: v_mov_b32_e32 v2, s10
; VGPRRC-NEXT: v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT: global_store_dwordx4 v[38:39], v[0:3], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: s_nop 0
; VGPRRC-NEXT: v_mov_b32_e32 v0, s12
; VGPRRC-NEXT: v_mov_b32_e32 v1, s13
; VGPRRC-NEXT: v_mov_b32_e32 v2, s14
; VGPRRC-NEXT: v_mov_b32_e32 v3, s15
-; VGPRRC-NEXT: global_store_dwordx4 v[36:37], v[0:3], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: s_endpgm
; AGPR-LABEL: test_mfma_i32_32x32x32_i8:
@@ -3386,47 +3398,47 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
; SDAG-NEXT: v_mov_b64_e32 v[0:1], 48
; SDAG-NEXT: v_mov_b64_e32 v[2:3], 32
-; SDAG-NEXT: v_mov_b64_e32 v[4:5], 16
; SDAG-NEXT: s_waitcnt lgkmcnt(0)
-; SDAG-NEXT: v_mov_b32_e32 v8, s24
-; SDAG-NEXT: v_mov_b32_e32 v9, s25
-; SDAG-NEXT: v_mov_b32_e32 v10, s26
-; SDAG-NEXT: v_mov_b32_e32 v11, s27
-; SDAG-NEXT: v_accvgpr_write_b32 a0, s8
-; SDAG-NEXT: v_mov_b32_e32 v12, s28
-; SDAG-NEXT: v_mov_b32_e32 v13, s29
-; SDAG-NEXT: v_mov_b32_e32 v14, s30
-; SDAG-NEXT: v_mov_b32_e32 v15, s31
-; SDAG-NEXT: v_accvgpr_write_b32 a1, s9
-; SDAG-NEXT: v_accvgpr_write_b32 a2, s10
-; SDAG-NEXT: v_accvgpr_write_b32 a3, s11
-; SDAG-NEXT: v_accvgpr_write_b32 a4, s12
-; SDAG-NEXT: v_accvgpr_write_b32 a5, s13
-; SDAG-NEXT: v_accvgpr_write_b32 a6, s14
-; SDAG-NEXT: v_accvgpr_write_b32 a7, s15
-; SDAG-NEXT: v_accvgpr_write_b32 a8, s16
-; SDAG-NEXT: v_accvgpr_write_b32 a9, s17
-; SDAG-NEXT: v_accvgpr_write_b32 a10, s18
-; SDAG-NEXT: v_accvgpr_write_b32 a11, s19
-; SDAG-NEXT: v_accvgpr_write_b32 a12, s20
-; SDAG-NEXT: v_accvgpr_write_b32 a13, s21
-; SDAG-NEXT: v_accvgpr_write_b32 a14, s22
-; SDAG-NEXT: v_accvgpr_write_b32 a15, s23
+; SDAG-NEXT: v_mov_b32_e32 v4, s24
+; SDAG-NEXT: v_mov_b32_e32 v5, s25
+; SDAG-NEXT: v_mov_b32_e32 v6, s26
+; SDAG-NEXT: v_mov_b32_e32 v7, s27
+; SDAG-NEXT: v_accvgpr_write_b32 a31, s23
+; SDAG-NEXT: v_mov_b32_e32 v8, s28
+; SDAG-NEXT: v_mov_b32_e32 v9, s29
+; SDAG-NEXT: v_mov_b32_e32 v10, s30
+; SDAG-NEXT: v_mov_b32_e32 v11, s31
+; SDAG-NEXT: v_accvgpr_write_b32 a30, s22
+; SDAG-NEXT: v_accvgpr_write_b32 a29, s21
+; SDAG-NEXT: v_accvgpr_write_b32 a28, s20
+; SDAG-NEXT: v_accvgpr_write_b32 a27, s19
+; SDAG-NEXT: v_accvgpr_write_b32 a26, s18
+; SDAG-NEXT: v_accvgpr_write_b32 a25, s17
+; SDAG-NEXT: v_accvgpr_write_b32 a24, s16
+; SDAG-NEXT: v_accvgpr_write_b32 a23, s15
+; SDAG-NEXT: v_accvgpr_write_b32 a22, s14
+; SDAG-NEXT: v_accvgpr_write_b32 a21, s13
+; SDAG-NEXT: v_accvgpr_write_b32 a20, s12
+; SDAG-NEXT: v_accvgpr_write_b32 a19, s11
+; SDAG-NEXT: v_accvgpr_write_b32 a18, s10
+; SDAG-NEXT: v_accvgpr_write_b32 a17, s9
+; SDAG-NEXT: v_accvgpr_write_b32 a16, s8
+; SDAG-NEXT: s_nop 1
+; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[4:7], v[8:11], a[16:31] cbsz:2 abid:3 blgp:1
+; SDAG-NEXT: v_mov_b64_e32 v[4:5], 16
; SDAG-NEXT: v_mov_b64_e32 v[6:7], 0
-; SDAG-NEXT: s_nop 0
-; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15] cbsz:2 abid:3 blgp:1
; SDAG-NEXT: v_mov_b32_e32 v8, s16
; SDAG-NEXT: v_mov_b32_e32 v9, s17
; SDAG-NEXT: v_mov_b32_e32 v10, s18
; SDAG-NEXT: v_mov_b32_e32 v11, s19
-; SDAG-NEXT: s_nop 7
-; SDAG-NEXT: global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
+; SDAG-NEXT: s_nop 5
+; SDAG-NEXT: global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
-; SDAG-NEXT: global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
+; SDAG-NEXT: global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
-; SDAG-NEXT: global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
+; SDAG-NEXT: global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
-; SDAG-NEXT: global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
+; SDAG-NEXT: global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
; SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
; SDAG-NEXT: s_waitcnt vmcnt(0)
@@ -3455,9 +3467,9 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
; GISEL: ; %bb.0:
; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24
; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
-; GISEL-NEXT: v_mov_b64_e32 v[20:21], 0
-; GISEL-NEXT: v_mov_b64_e32 v[26:27], 48
-; GISEL-NEXT: v_mov_b64_e32 v[22:23], 16
+; GISEL-NEXT: v_mov_b64_e32 v[12:13], 0
+; GISEL-NEXT: v_mov_b64_e32 v[14:15], 16
+; GISEL-NEXT: v_mov_b64_e32 v[16:17], 32
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25]
; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27]
@@ -3480,31 +3492,34 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
; GISEL-NEXT: v_accvgpr_write_b32 a14, s22
; GISEL-NEXT: v_accvgpr_write_b32 a15, s23
; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9]
-; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[18:19], 48
; GISEL-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
-; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
-; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17]
-; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
-; GISEL-NEXT: v_mov_b64_e32 v[24:25], 32
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[12:13]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[14:15]
; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11]
-; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15]
-; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19]
-; GISEL-NEXT: s_nop 4
-; GISEL-NEXT: global_store_dwordx4 v[20:21], a[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 7
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: global_store_dwordx4 v[12:13], a[16:19], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], a[20:23], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], a[20:23], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], a[24:27], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[16:17], a[24:27], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], a[28:31], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[18:19], a[28:31], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[20:21], v[8:11], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[12:13], v[8:11], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[22:23], v[12:15], off sc0 sc1
+; GISEL-NEXT: global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[24:25], v[16:19], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[16:17]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[18:19]
+; GISEL-NEXT: global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
-; GISEL-NEXT: global_store_dwordx4 v[26:27], v[0:3], off sc0 sc1
+; GISEL-NEXT: s_nop 0
+; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
+; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
+; GISEL-NEXT: global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
; GISEL-NEXT: s_waitcnt vmcnt(0)
; GISEL-NEXT: s_endpgm
;
@@ -3514,47 +3529,47 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
; HEURRC-NEXT: v_mov_b64_e32 v[0:1], 48
; HEURRC-NEXT: v_mov_b64_e32 v[2:3], 32
-; HEURRC-NEXT: v_mov_b64_e32 v[4:5], 16
; HEURRC-NEXT: s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT: v_mov_b32_e32 v8, s24
-; HEURRC-NEXT: v_mov_b32_e32 v9, s25
-; HEURRC-NEXT: v_mov_b32_e32 v10, s26
-; HEURRC-NEXT: v_mov_b32_e32 v11, s27
-; HEURRC-NEXT: v_accvgpr_write_b32 a0, s8
-; HEURRC-NEXT: v_mov_b32_e32 v12, s28
-; HEURRC-NEXT: v_mov_b32_e32 v13, s29
-; HEURRC-NEXT: v_mov_b32_e32 v14, s30
-; HEURRC-NEXT: v_mov_b32_e32 v15, s31
-; HEURRC-NEXT: v_accvgpr_write_b32 a1, s9
-; HEURRC-NEXT: v_accvgpr_write_b32 a2, s10
-; HEURRC-NEXT: v_accvgpr_write_b32 a3, s11
-; HEURRC-NEXT: v_accvgpr_write_b32 a4, s12
-; HEURRC-NEXT: v_accvgpr_write_b32 a5, s13
-; HEURRC-NEXT: v_accvgpr_write_b32 a6, s14
-; HEURRC-NEXT: v_accvgpr_write_b32 a7, s15
-; HEURRC-NEXT: v_accvgpr_write_b32 a8, s16
-; HEURRC-NEXT: v_accvgpr_write_b32 a9, s17
-; HEURRC-NEXT: v_accvgpr_write_b32 a10, s18
-; HEURRC-NEXT: v_accvgpr_write_b32 a11, s19
-; HEURRC-NEXT: v_accvgpr_write_b32 a12, s20
-; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21
-; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22
-; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23
+; HEURRC-NEXT: v_mov_b32_e32 v4, s24
+; HEURRC-NEXT: v_mov_b32_e32 v5, s25
+; HEURRC-NEXT: v_mov_b32_e32 v6, s26
+; HEURRC-NEXT: v_mov_b32_e32 v7, s27
+; HEURRC-NEXT: v_accvgpr_write_b32 a31, s23
+; HEURRC-NEXT: v_mov_b32_e32 v8, s28
+; HEURRC-NEXT: v_mov_b32_e32 v9, s29
+; HEURRC-NEXT: v_mov_b32_e32 v10, s30
+; HEURRC-NEXT: v_mov_b32_e32 v11, s31
+; HEURRC-NEXT: v_accvgpr_write_b32 a30, s22
+; HEURRC-NEXT: v_accvgpr_write_b32 a29, s21
+; HEURRC-NEXT: v_accvgpr_write_b32 a28, s20
+; HEURRC-NEXT: v_accvgpr_write_b32 a27, s19
+; HEURRC-NEXT: v_accvgpr_write_b32 a26, s18
+; HEURRC-NEXT: v_accvgpr_write_b32 a25, s17
+; HEURRC-NEXT: v_accvgpr_write_b32 a24, s16
+; HEURRC-NEXT: v_accvgpr_write_b32 a23, s15
+; HEURRC-NEXT: v_accvgpr_write_b32 a22, s14
+; HEURRC-NEXT: v_accvgpr_write_b32 a21, s13
+; HEURRC-NEXT: v_accvgpr_write_b32 a20, s12
+; HEURRC-NEXT: v_accvgpr_write_b32 a19, s11
+; HEURRC-NEXT: v_accvgpr_write_b32 a18, s10
+; HEURRC-NEXT: v_accvgpr_write_b32 a17, s9
+; HEURRC-NEXT: v_accvgpr_write_b32 a16, s8
+; HEURRC-NEXT: s_nop 1
+; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[4:7], v[8:11], a[16:31] cbsz:2 abid:3 blgp:1
+; HEURRC-NEXT: v_mov_b64_e32 v[4:5], 16
; HEURRC-NEXT: v_mov_b64_e32 v[6:7], 0
-; HEURRC-NEXT: s_nop 0
-; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15] cbsz:2 abid:3 blgp:1
; HEURRC-NEXT: v_mov_b32_e32 v8, s16
; HEURRC-NEXT: v_mov_b32_e32 v9, s17
; HEURRC-NEXT: v_mov_b32_e32 v10, s18
; HEURRC-NEXT: v_mov_b32_e32 v11, s19
-; HEURRC-NEXT: s_nop 7
-; HEURRC-NEXT: global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
+; HEURRC-NEXT: s_nop 5
+; HEURRC-NEXT: global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
-; HEURRC-NEXT: global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
+; HEURRC-NEXT: global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
-; HEURRC-NEXT: global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
+; HEURRC-NEXT: global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
-; HEURRC-NEXT: global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
+; HEURRC-NEXT: global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
; HEURRC-NEXT: global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
; HEURRC-NEXT: s_waitcnt vmcnt(0)
@@ -3585,37 +3600,40 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
; VGPRRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64
; VGPRRC-NEXT: v_mov_b64_e32 v[32:33], 48
; VGPRRC-NEXT: v_mov_b64_e32 v[34:35], 32
-; VGPRRC-NEXT: v_mov_b64_e32 v[36:37], 16
; VGPRRC-NEXT: s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT: v_mov_b32_e32 v40, s24
-; VGPRRC-NEXT: v_mov_b32_e32 v41, s25
-; VGPRRC-NEXT: v_mov_b32_e32 v42, s26
-; VGPRRC-NEXT: v_mov_b32_e32 v43, s27
-; VGPRRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9]
-; VGPRRC-NEXT: v_mov_b32_e32 v44, s28
-; VGPRRC-NEXT: v_mov_b32_e32 v45, s29
-; VGPRRC-NEXT: v_mov_b32_e32 v46, s30
-; VGPRRC-NEXT: v_mov_b32_e32 v47, s31
-; VGPRRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11]
-; VGPRRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13]
-; VGPRRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15]
-; VGPRRC-NEXT: v_mov_b64_e32 v[8:9], s[16:17]
-; VGPRRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19]
-; VGPRRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21]
-; VGPRRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT: v_mov_b64_e32 v[38:39], 0
-; VGPRRC-NEXT: s_nop 0
-; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[16:31], v[40:43], v[44:47], v[0:15] cbsz:2 abid:3 blgp:1
+; VGPRRC-NEXT: v_mov_b32_e32 v36, s24
+; VGPRRC-NEXT: v_mov_b32_e32 v37, s25
+; VGPRRC-NEXT: v_mov_b32_e32 v38, s26
+; VGPRRC-NEXT: v_mov_b32_e32 v39, s27
+; VGPRRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23]
+; VGPRRC-NEXT: v_mov_b32_e32 v40, s28
+; VGPRRC-NEXT: v_mov_b32_e32 v41, s29
+; VGPRRC-NEXT: v_mov_b32_e32 v42, s30
+; VGPRRC-NEXT: v_mov_b32_e32 v43, s31
+; VGPRRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21]
+; VGPRRC-NEXT: v_mov_b64_e32 v[26:27], s[18:19]
+; VGPRRC-NEXT: v_mov_b64_e32 v[24:25], s[16:17]
+; VGPRRC-NEXT: v_mov_b64_e32 v[22:23], s[14:15]
+; VGPRRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13]
+; VGPRRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11]
+; VGPRRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9]
+; VGPRRC-NEXT: s_nop 1
+; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[36:39], v[40:43], v[16:31] cbsz:2 abid:3 blgp:1
; VGPRRC-NEXT: s_nop 7
; VGPRRC-NEXT: s_nop 3
-; VGPRRC-NEXT: global_store_dwordx4 v[32:33], v[28:31], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[32:33], v[12:15], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[34:35], v[24:27], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[34:35], v[8:11], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[36:37], v[20:23], off sc0 sc1
+; VGPRRC-NEXT: s_nop 0
+; VGPRRC-NEXT: v_mov_b64_e32 v[8:9], 16
+; VGPRRC-NEXT: global_store_dwordx4 v[8:9], v[4:7], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
-; VGPRRC-NEXT: global_store_dwordx4 v[38:39], v[16:19], off sc0 sc1
+; VGPRRC-NEXT: s_nop 0
+; VGPRRC-NEXT: v_mov_b64_e32 v[4:5], 0
+; VGPRRC-NEXT: global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
+; VGPRRC-NEXT: s_nop 0
; VGPRRC-NEXT: v_mov_b32_e32 v0, s16
; VGPRRC-NEXT: v_mov_b32_e32 v1, s17
; VGPRRC-NEXT: v_mov_b32_e32 v2, s18
@@ -3634,14 +3652,14 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
; VGPRRC-NEXT: v_mov_b32_e32 v1, s9
; VGPRRC-NEXT: v_mov_b32_e32 v2, s10
; VGPRRC-NEXT: v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT: global_store_dwordx4 v[38:39], v[0:3], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: s_nop 0
; VGPRRC-NEXT: v_mov_b32_e32 v0, s12
; VGPRRC-NEXT: v_mov_b32_e32 v1, s13
; VGPRRC-NEXT: v_mov_b32_e32 v2, s14
; VGPRRC-NEXT: v_mov_b32_e32 v3, s15
-; VGPRRC-NEXT: global_store_dwordx4 v[36:37], v[0:3], off sc0 sc1
+; VGPRRC-NEXT: global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
; VGPRRC-NEXT: s_waitcnt vmcnt(0)
; VGPRRC-NEXT: s_endpgm
; AGPR-LABEL: test_mfma_i32_32x32x32_i8__flags:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll
index 0b2818f38149d..24af3fa5ff9b7 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll
@@ -4784,8 +4784,8 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__nonmac(<8 x
; GISEL-NEXT: s_load_dwordx16 s[36:51], s[4:5], 0x0
; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x40
; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x80
+; GISEL-NEXT: v_mov_b64_e32 v[16:17], 0
; GISEL-NEXT: v_mov_b64_e32 v[18:19], 16
-; GISEL-NEXT: v_mov_b64_e32 v[20:21], 32
; GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[36:37]
; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[38:39]
@@ -4811,16 +4811,16 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__nonmac(<8 x
; GISEL-NEXT: v_accvgpr_write_b32 a13, s21
; GISEL-NEXT: v_accvgpr_write_b32 a14, s22
; GISEL-NEXT: v_accvgpr_write_b32 a15, s23
-; GISEL-NEXT: v_mov_b32_e32 v16, s1
+; GISEL-NEXT: v_mov_b32_e32 v20, s1
; GISEL-NEXT: v_mov_b64_e32 v[22:23], 48
; GISEL-NEXT: s_nop 0
-; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], s0, v16 op_sel_hi:[0,0,0]
+; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], s0, v20 op_sel_hi:[0,0,0]
; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9]
-; GISEL-NEXT: v_mov_b64_e32 v[16:17], 0
; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11]
; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13]
; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17]
; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21]
+; GISEL-NEXT: v_mov_b64_e32 v[20:21], 32
; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15]
; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19]
; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23]
More information about the llvm-branch-commits
mailing list