[llvm] [AMDGPU] Avoid put implicit_def into bundle that break reg's liveness (PR #142563)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Jun 11 21:41:35 PDT 2025
https://github.com/Shoreshen updated https://github.com/llvm/llvm-project/pull/142563
>From 3dc03fba1f04940d226459ca41dfc33573efa072 Mon Sep 17 00:00:00 2001
From: shore <372660931 at qq.com>
Date: Tue, 3 Jun 2025 17:11:27 +0800
Subject: [PATCH 1/2] avoid put implicit_def into bundle that break reg's
liveness
---
.../lib/Target/AMDGPU/SIInsertHardClauses.cpp | 2 +-
llvm/lib/Target/AMDGPU/SIPostRABundler.cpp | 3 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll | 940 +++++++++---------
.../CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll | 46 +-
.../AMDGPU/bundle-break-phy-liveness.mir | 31 +
5 files changed, 532 insertions(+), 490 deletions(-)
create mode 100644 llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir
diff --git a/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp b/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp
index eb0977b92d5ab..1f3e549b0e27f 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp
@@ -145,7 +145,7 @@ class SIInsertHardClauses {
// It's safe to treat the rest as illegal.
if (MI.getOpcode() == AMDGPU::S_NOP)
return HARDCLAUSE_INTERNAL;
- if (MI.isMetaInstruction())
+ if (MI.isMetaInstruction() && MI.getOpcode() != AMDGPU::IMPLICIT_DEF)
return HARDCLAUSE_IGNORE;
return HARDCLAUSE_ILLEGAL;
}
diff --git a/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp b/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp
index efdc55b8e68be..48f84286e9214 100644
--- a/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp
+++ b/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp
@@ -184,7 +184,8 @@ bool SIPostRABundler::run(MachineFunction &MF) {
if (I->getNumExplicitDefs() != 0)
Defs.insert(I->defs().begin()->getReg());
++ClauseLength;
- } else if (!I->isMetaInstruction()) {
+ } else if (!I->isMetaInstruction() ||
+ I->getOpcode() == AMDGPU::IMPLICIT_DEF) {
// Allow meta instructions in between bundle candidates, but do not
// start or end a bundle on one.
//
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
index 44abfd272be88..1ab1b5b3b202d 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
@@ -2,11 +2,11 @@
; FIXME: Currently block machineinstr verifier due to SI BUNDLE pass break physical register liveness. Should remove when the issue is fixed up
-; RUN: llc -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs=0 < %s | FileCheck -check-prefix=SI %s
-; RUN: llc -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs=0 < %s | FileCheck -check-prefix=VI %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs=0 < %s | FileCheck -check-prefix=GFX9 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -verify-machineinstrs=0 < %s | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -verify-machineinstrs=0 < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
+; RUN: llc -mtriple=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -mtriple=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
define <32 x float> @bitcast_v32i32_to_v32f32(<32 x i32> %a, i32 %b) {
; SI-LABEL: bitcast_v32i32_to_v32f32:
@@ -4334,9 +4334,14 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr59
+; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
@@ -4438,18 +4443,45 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr59
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr53
+; VI-NEXT: ; implicit-def: $vgpr57
+; VI-NEXT: ; implicit-def: $vgpr56
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr52
+; VI-NEXT: ; implicit-def: $vgpr47
+; VI-NEXT: ; implicit-def: $vgpr46
+; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr51
+; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: ; implicit-def: $vgpr63
+; VI-NEXT: ; implicit-def: $vgpr36
+; VI-NEXT: ; implicit-def: $vgpr62
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; implicit-def: $vgpr60
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr41
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr58
@@ -4479,43 +4511,12 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr53
-; VI-NEXT: ; implicit-def: $vgpr57
-; VI-NEXT: ; implicit-def: $vgpr56
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr47
-; VI-NEXT: ; implicit-def: $vgpr46
-; VI-NEXT: ; implicit-def: $vgpr45
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: ; implicit-def: $vgpr51
-; VI-NEXT: ; implicit-def: $vgpr37
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr63
-; VI-NEXT: ; implicit-def: $vgpr36
-; VI-NEXT: ; implicit-def: $vgpr62
-; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; implicit-def: $vgpr60
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: ; implicit-def: $vgpr58
-; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB12_2
; VI-NEXT: ; %bb.1: ; %cmp.false
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v32
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v31
@@ -4694,6 +4695,7 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB12_4
; VI-NEXT: ; %bb.3: ; %cmp.true
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_add_u32_e32 v32, vcc, 3, v32
; VI-NEXT: v_add_u32_e32 v31, vcc, 3, v31
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[31:32]
@@ -5293,9 +5295,16 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr40
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
@@ -5389,14 +5398,12 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; kill: killed $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr56
; GFX9-NEXT: ; implicit-def: $vgpr44
; GFX9-NEXT: ; implicit-def: $vgpr38
@@ -5428,15 +5435,15 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr42
+; GFX9-NEXT: s_waitcnt vmcnt(6)
+; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; GFX9-NEXT: ; implicit-def: $vgpr33
+; GFX9-NEXT: ; kill: killed $vgpr33
+; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
@@ -5484,11 +5491,6 @@ define <128 x i8> @bitcast_v32i32_to_v128i8(<32 x i32> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(30)
-; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB12_2
@@ -42059,9 +42061,14 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr59
+; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
@@ -42163,18 +42170,45 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr59
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr53
+; VI-NEXT: ; implicit-def: $vgpr57
+; VI-NEXT: ; implicit-def: $vgpr56
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr52
+; VI-NEXT: ; implicit-def: $vgpr47
+; VI-NEXT: ; implicit-def: $vgpr46
+; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr51
+; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: ; implicit-def: $vgpr63
+; VI-NEXT: ; implicit-def: $vgpr36
+; VI-NEXT: ; implicit-def: $vgpr62
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; implicit-def: $vgpr60
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr41
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr58
@@ -42204,43 +42238,12 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr53
-; VI-NEXT: ; implicit-def: $vgpr57
-; VI-NEXT: ; implicit-def: $vgpr56
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr47
-; VI-NEXT: ; implicit-def: $vgpr46
-; VI-NEXT: ; implicit-def: $vgpr45
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: ; implicit-def: $vgpr51
-; VI-NEXT: ; implicit-def: $vgpr37
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr63
-; VI-NEXT: ; implicit-def: $vgpr36
-; VI-NEXT: ; implicit-def: $vgpr62
-; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; implicit-def: $vgpr60
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: ; implicit-def: $vgpr58
-; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB36_2
; VI-NEXT: ; %bb.1: ; %cmp.false
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v32
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v31
@@ -42419,6 +42422,7 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB36_4
; VI-NEXT: ; %bb.3: ; %cmp.true
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_add_f32_e32 v32, 1.0, v32
; VI-NEXT: v_add_f32_e32 v31, 1.0, v31
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[31:32]
@@ -43018,9 +43022,16 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr40
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
@@ -43114,14 +43125,12 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; kill: killed $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr56
; GFX9-NEXT: ; implicit-def: $vgpr44
; GFX9-NEXT: ; implicit-def: $vgpr38
@@ -43153,15 +43162,15 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr42
+; GFX9-NEXT: s_waitcnt vmcnt(6)
+; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; GFX9-NEXT: ; implicit-def: $vgpr33
+; GFX9-NEXT: ; kill: killed $vgpr33
+; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
@@ -43209,11 +43218,6 @@ define <128 x i8> @bitcast_v32f32_to_v128i8(<32 x float> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(30)
-; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB36_2
@@ -79755,9 +79759,14 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr59
+; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
@@ -79859,18 +79868,45 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr59
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr53
+; VI-NEXT: ; implicit-def: $vgpr57
+; VI-NEXT: ; implicit-def: $vgpr56
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr52
+; VI-NEXT: ; implicit-def: $vgpr47
+; VI-NEXT: ; implicit-def: $vgpr46
+; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr51
+; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: ; implicit-def: $vgpr63
+; VI-NEXT: ; implicit-def: $vgpr36
+; VI-NEXT: ; implicit-def: $vgpr62
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; implicit-def: $vgpr60
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr41
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr58
@@ -79900,43 +79936,12 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr53
-; VI-NEXT: ; implicit-def: $vgpr57
-; VI-NEXT: ; implicit-def: $vgpr56
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr47
-; VI-NEXT: ; implicit-def: $vgpr46
-; VI-NEXT: ; implicit-def: $vgpr45
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: ; implicit-def: $vgpr51
-; VI-NEXT: ; implicit-def: $vgpr37
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr63
-; VI-NEXT: ; implicit-def: $vgpr36
-; VI-NEXT: ; implicit-def: $vgpr62
-; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; implicit-def: $vgpr60
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: ; implicit-def: $vgpr58
-; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB56_2
; VI-NEXT: ; %bb.1: ; %cmp.false
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v32
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v31
@@ -80145,6 +80150,7 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; VI-NEXT: v_addc_u32_e32 v28, vcc, 0, v28, vcc
; VI-NEXT: v_add_u32_e32 v29, vcc, 3, v29
; VI-NEXT: v_addc_u32_e32 v30, vcc, 0, v30, vcc
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_add_u32_e32 v31, vcc, 3, v31
; VI-NEXT: v_addc_u32_e32 v32, vcc, 0, v32, vcc
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[31:32]
@@ -80714,9 +80720,16 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr40
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
@@ -80810,14 +80823,12 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; kill: killed $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr56
; GFX9-NEXT: ; implicit-def: $vgpr44
; GFX9-NEXT: ; implicit-def: $vgpr38
@@ -80849,15 +80860,15 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; GFX9-NEXT: ; kill: killed $vgpr40
; GFX9-NEXT: ; implicit-def: $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr42
+; GFX9-NEXT: s_waitcnt vmcnt(6)
+; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; GFX9-NEXT: ; implicit-def: $vgpr33
+; GFX9-NEXT: ; kill: killed $vgpr33
+; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
@@ -80905,11 +80916,6 @@ define <128 x i8> @bitcast_v16i64_to_v128i8(<16 x i64> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(30)
-; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB56_2
@@ -115517,9 +115523,14 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr57
+; VI-NEXT: ; implicit-def: $vgpr56
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr56
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
@@ -115619,18 +115630,45 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr57
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr56
; VI-NEXT: ; kill: killed $vgpr39
; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr56
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr47
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr46
+; VI-NEXT: ; implicit-def: $vgpr53
+; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr52
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr63
+; VI-NEXT: ; implicit-def: $vgpr62
+; VI-NEXT: ; implicit-def: $vgpr51
+; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: ; implicit-def: $vgpr60
+; VI-NEXT: ; implicit-def: $vgpr59
+; VI-NEXT: ; implicit-def: $vgpr36
+; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; kill: killed $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr41
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr56
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr56
@@ -115663,43 +115701,12 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr56
; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr47
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr46
-; VI-NEXT: ; implicit-def: $vgpr53
-; VI-NEXT: ; implicit-def: $vgpr45
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr63
-; VI-NEXT: ; implicit-def: $vgpr62
-; VI-NEXT: ; implicit-def: $vgpr51
-; VI-NEXT: ; implicit-def: $vgpr37
-; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr60
-; VI-NEXT: ; implicit-def: $vgpr59
-; VI-NEXT: ; implicit-def: $vgpr36
-; VI-NEXT: ; implicit-def: $vgpr58
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; kill: killed $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; implicit-def: $vgpr56
-; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB72_2
; VI-NEXT: ; %bb.1: ; %cmp.false
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v32
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v32
@@ -115878,6 +115885,7 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB72_4
; VI-NEXT: ; %bb.3: ; %cmp.true
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_add_f64 v[31:32], v[31:32], 1.0
; VI-NEXT: v_add_f64 v[29:30], v[29:30], 1.0
; VI-NEXT: v_add_f64 v[27:28], v[27:28], 1.0
@@ -116464,9 +116472,16 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr41
; GFX9-NEXT: ; kill: killed $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr41
@@ -116560,14 +116575,12 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr48
; GFX9-NEXT: ; kill: killed $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr41
-; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; kill: killed $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; kill: killed $vgpr48
; GFX9-NEXT: ; implicit-def: $vgpr48
; GFX9-NEXT: ; kill: killed $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr41
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr46
; GFX9-NEXT: ; implicit-def: $vgpr54
; GFX9-NEXT: ; implicit-def: $vgpr45
@@ -116599,15 +116612,13 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr48
; GFX9-NEXT: ; kill: killed $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr41
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr42
+; GFX9-NEXT: s_waitcnt vmcnt(6)
+; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
@@ -116659,9 +116670,6 @@ define <128 x i8> @bitcast_v16f64_to_v128i8(<16 x double> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(32)
-; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
-; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB72_2
@@ -164579,12 +164587,16 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr34
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr44
; VI-NEXT: ; implicit-def: $vgpr38
; VI-NEXT: ; implicit-def: $vgpr55
; VI-NEXT: ; implicit-def: $vgpr54
@@ -164609,52 +164621,11 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: ; implicit-def: $vgpr52
; VI-NEXT: ; implicit-def: $vgpr37
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr56
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr56
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(14)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: ; kill: killed $vgpr33
@@ -164763,6 +164734,43 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: ; kill: killed $vgpr33
; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB90_2
@@ -164897,6 +164905,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[1:2]
; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[31:32]
; VI-NEXT: v_lshrrev_b32_e32 v46, 24, v12
; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
@@ -166116,10 +166125,60 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr59
+; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr59
+; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr59
+; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; kill: killed $vgpr59
+; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; kill: killed $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: v_mov_b32_e32 v46, v15
; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: ; kill: killed $vgpr59
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
@@ -166129,86 +166188,42 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: v_mov_b32_e32 v47, v16
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr36
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; implicit-def: $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr38
; GFX9-NEXT: ; implicit-def: $vgpr51
@@ -166239,15 +166254,25 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr57
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr63
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: s_waitcnt vmcnt(6)
+; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
@@ -166303,21 +166328,6 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(34)
-; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB90_2
@@ -192601,9 +192611,12 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v27
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v25
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v23
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
; VI-NEXT: v_lshrrev_b32_e32 v48, 16, v16
; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v15
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
; VI-NEXT: v_lshrrev_b32_e32 v54, 16, v12
; VI-NEXT: v_lshrrev_b32_e32 v62, 16, v11
; VI-NEXT: v_lshrrev_b32_e32 v63, 16, v10
@@ -192613,19 +192626,20 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v2
; VI-NEXT: v_lshrrev_b32_e32 v56, 16, v28
; VI-NEXT: v_lshrrev_b32_e32 v58, 16, v26
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v24
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v23
; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v22
; VI-NEXT: v_lshrrev_b32_e32 v38, 16, v21
; VI-NEXT: v_lshrrev_b32_e32 v53, 16, v20
; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v19
; VI-NEXT: v_lshrrev_b32_e32 v41, 16, v18
; VI-NEXT: v_lshrrev_b32_e32 v46, 16, v17
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr52
; VI-NEXT: ; implicit-def: $vgpr45
; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; implicit-def: $vgpr35
; VI-NEXT: s_waitcnt vmcnt(13)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; VI-NEXT: ; implicit-def: $vgpr31
@@ -192723,15 +192737,15 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: s_waitcnt vmcnt(12)
-; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v61
-; VI-NEXT: s_waitcnt vmcnt(11)
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v60
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr31
+; VI-NEXT: s_waitcnt vmcnt(14)
+; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v61
+; VI-NEXT: s_waitcnt vmcnt(13)
+; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v60
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr31
@@ -192743,15 +192757,11 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; implicit-def: $vgpr35
; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr32
+; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr32
@@ -193648,9 +193658,17 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr53
; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr44
+; GFX9-NEXT: ; implicit-def: $vgpr53
+; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
@@ -193710,10 +193728,8 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: ; implicit-def: $vgpr44
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: ; implicit-def: $vgpr53
; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr57
@@ -193740,41 +193756,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr52
; GFX9-NEXT: ; implicit-def: $vgpr51
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr53
; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr54
-; GFX9-NEXT: ; implicit-def: $vgpr53
-; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(18)
+; GFX9-NEXT: s_waitcnt vmcnt(8)
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; kill: killed $vgpr33
@@ -193833,6 +193822,27 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
+; GFX9-NEXT: ; implicit-def: $vgpr54
+; GFX9-NEXT: ; implicit-def: $vgpr53
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB94_2
@@ -216309,12 +216319,74 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v27
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v26
+; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v25
+; VI-NEXT: ; kill: killed $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v24
+; VI-NEXT: ; kill: killed $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v23
+; VI-NEXT: ; kill: killed $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v22
+; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v21
+; VI-NEXT: ; kill: killed $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; kill: killed $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
; VI-NEXT: v_lshrrev_b32_e32 v60, 16, v16
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
; VI-NEXT: v_lshrrev_b32_e32 v63, 16, v12
@@ -216324,20 +216396,34 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v56, 16, v4
; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v2
; VI-NEXT: v_lshrrev_b32_e32 v46, 16, v30
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v22
-; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v21
; VI-NEXT: v_lshrrev_b32_e32 v53, 16, v20
; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v19
; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v18
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr32
+; VI-NEXT: ; implicit-def: $vgpr62
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: ; implicit-def: $vgpr39
+; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr41
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v37
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v50, 16, v36
; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
@@ -216408,7 +216494,6 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
@@ -216418,88 +216503,17 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
+; VI-NEXT: v_lshrrev_b32_e32 v50, 16, v36
; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr55
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr62
; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr58
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr54
@@ -216508,14 +216522,10 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr45
-; VI-NEXT: ; implicit-def: $vgpr44
-; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr48
@@ -217510,9 +217520,17 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr53
; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr44
+; GFX9-NEXT: ; implicit-def: $vgpr53
+; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
@@ -217572,10 +217590,8 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: ; implicit-def: $vgpr44
; GFX9-NEXT: ; kill: killed $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: ; implicit-def: $vgpr53
; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr57
@@ -217602,41 +217618,14 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr52
; GFX9-NEXT: ; implicit-def: $vgpr51
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr53
; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr54
-; GFX9-NEXT: ; implicit-def: $vgpr53
-; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(18)
+; GFX9-NEXT: s_waitcnt vmcnt(8)
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; kill: killed $vgpr33
@@ -217695,6 +217684,27 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; kill: killed $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr33
+; GFX9-NEXT: ; implicit-def: $vgpr54
+; GFX9-NEXT: ; implicit-def: $vgpr53
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB98_2
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
index 397955a8a8928..dc53b27bc95a8 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
@@ -65436,8 +65436,28 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
;
; VI-LABEL: bitcast_v32i16_to_v64i8:
; VI: ; %bb.0:
-; VI-NEXT: ; implicit-def: $vgpr19
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr19
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v17
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
@@ -65458,23 +65478,6 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr43
; VI-NEXT: ; kill: killed $vgpr17
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
@@ -65500,11 +65503,6 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr41
; VI-NEXT: ; implicit-def: $vgpr18
; VI-NEXT: ; implicit-def: $vgpr54
@@ -65536,6 +65534,8 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr21
; VI-NEXT: ; implicit-def: $vgpr20
; VI-NEXT: ; implicit-def: $vgpr19
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr42
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
diff --git a/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir b/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir
new file mode 100644
index 0000000000000..57476be8bc3c3
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir
@@ -0,0 +1,31 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -run-pass=si-insert-hard-clauses -verify-machineinstrs %s -o - | FileCheck -check-prefix=GCN-CLAUSE %s
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -run-pass=si-post-ra-bundler -verify-machineinstrs %s -o - | FileCheck -check-prefix=GCN-BUNDLE %s
+
+---
+name: clause_implicit_def
+tracksRegLiveness: true
+body: |
+ bb.0:
+ liveins: $vgpr0, $sgpr32
+
+ ; GCN-CLAUSE-LABEL: name: clause_implicit_def
+ ; GCN-CLAUSE: liveins: $vgpr0, $sgpr32
+ ; GCN-CLAUSE-NEXT: {{ $}}
+ ; GCN-CLAUSE-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
+ ; GCN-CLAUSE-NEXT: $vgpr2 = IMPLICIT_DEF
+ ; GCN-CLAUSE-NEXT: $vgpr3 = IMPLICIT_DEF
+ ; GCN-CLAUSE-NEXT: SCRATCH_STORE_DWORDX2_SADDR $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+ ;
+ ; GCN-BUNDLE-LABEL: name: clause_implicit_def
+ ; GCN-BUNDLE: liveins: $vgpr0, $sgpr32
+ ; GCN-BUNDLE-NEXT: {{ $}}
+ ; GCN-BUNDLE-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
+ ; GCN-BUNDLE-NEXT: $vgpr2 = IMPLICIT_DEF
+ ; GCN-BUNDLE-NEXT: $vgpr3 = IMPLICIT_DEF
+ ; GCN-BUNDLE-NEXT: SCRATCH_STORE_DWORDX2_SADDR $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+ SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
+ $vgpr2 = IMPLICIT_DEF
+ $vgpr3 = IMPLICIT_DEF
+ SCRATCH_STORE_DWORDX2_SADDR $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+...
>From f3550df6a05a7ff8bb43d21e6d4bc75cbae8a85a Mon Sep 17 00:00:00 2001
From: shore <372660931 at qq.com>
Date: Fri, 6 Jun 2025 12:01:42 +0800
Subject: [PATCH 2/2] Applying comments from Jayfold and Matthew, using
finilizebundle to fix problem
---
llvm/lib/CodeGen/MachineInstrBundle.cpp | 17 ++++++-
.../lib/Target/AMDGPU/SIInsertHardClauses.cpp | 2 +-
llvm/lib/Target/AMDGPU/SIPostRABundler.cpp | 3 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll | 46 +++++++++----------
.../AMDGPU/bundle-break-phy-liveness.mir | 21 +++++----
5 files changed, 54 insertions(+), 35 deletions(-)
diff --git a/llvm/lib/CodeGen/MachineInstrBundle.cpp b/llvm/lib/CodeGen/MachineInstrBundle.cpp
index 34896c67144bc..d23acd25ee85d 100644
--- a/llvm/lib/CodeGen/MachineInstrBundle.cpp
+++ b/llvm/lib/CodeGen/MachineInstrBundle.cpp
@@ -115,6 +115,21 @@ static DebugLoc getDebugLoc(MachineBasicBlock::instr_iterator FirstMI,
return DebugLoc();
}
+static bool containRegOrSubReg(SmallSetVector<Register, 32> DefRegs,
+ Register Reg, const TargetRegisterInfo *TRI) {
+ if (DefRegs.contains(Reg))
+ return true;
+ if (Reg.isPhysical()) {
+ for (const MCPhysReg &SubReg : TRI->subregs(Reg)) {
+ // Applying same logic with MachineVerifier that any of the SubReg is
+ // contained, it counts defined
+ if (DefRegs.contains(SubReg))
+ return true;
+ }
+ }
+ return false;
+}
+
/// finalizeBundle - Finalize a machine instruction bundle which includes
/// a sequence of instructions starting from FirstMI to LastMI (exclusive).
/// This routine adds a BUNDLE instruction to represent the bundle, it adds
@@ -151,7 +166,7 @@ void llvm::finalizeBundle(MachineBasicBlock &MBB,
if (!Reg)
continue;
- if (LocalDefs.contains(Reg)) {
+ if (containRegOrSubReg(LocalDefs, Reg, TRI)) {
MO.setIsInternalRead();
if (MO.isKill()) {
// Internal def is now killed.
diff --git a/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp b/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp
index bdd4f8bc321b5..d8fe8505bc722 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertHardClauses.cpp
@@ -151,7 +151,7 @@ class SIInsertHardClauses {
// It's safe to treat the rest as illegal.
if (MI.getOpcode() == AMDGPU::S_NOP)
return HARDCLAUSE_INTERNAL;
- if (MI.isMetaInstruction() && MI.getOpcode() != AMDGPU::IMPLICIT_DEF)
+ if (MI.isMetaInstruction())
return HARDCLAUSE_IGNORE;
return HARDCLAUSE_ILLEGAL;
}
diff --git a/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp b/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp
index 48f84286e9214..efdc55b8e68be 100644
--- a/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp
+++ b/llvm/lib/Target/AMDGPU/SIPostRABundler.cpp
@@ -184,8 +184,7 @@ bool SIPostRABundler::run(MachineFunction &MF) {
if (I->getNumExplicitDefs() != 0)
Defs.insert(I->defs().begin()->getReg());
++ClauseLength;
- } else if (!I->isMetaInstruction() ||
- I->getOpcode() == AMDGPU::IMPLICIT_DEF) {
+ } else if (!I->isMetaInstruction()) {
// Allow meta instructions in between bundle candidates, but do not
// start or end a bundle on one.
//
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
index 4bd36705003e8..1024c2a7f066a 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
@@ -65448,28 +65448,8 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
;
; VI-LABEL: bitcast_v32i16_to_v64i8:
; VI: ; %bb.0:
-; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr19
+; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v17
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
@@ -65490,6 +65470,23 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
; VI-NEXT: ; implicit-def: $vgpr19
+; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr43
; VI-NEXT: ; kill: killed $vgpr17
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
@@ -65515,6 +65512,11 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; kill: killed $vgpr19
; VI-NEXT: ; implicit-def: $vgpr19
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr41
; VI-NEXT: ; implicit-def: $vgpr18
; VI-NEXT: ; implicit-def: $vgpr54
@@ -65546,8 +65548,6 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr21
; VI-NEXT: ; implicit-def: $vgpr20
; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr42
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
diff --git a/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir b/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir
index 57476be8bc3c3..8045fafc6a6c4 100644
--- a/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir
+++ b/llvm/test/CodeGen/AMDGPU/bundle-break-phy-liveness.mir
@@ -12,18 +12,23 @@ body: |
; GCN-CLAUSE-LABEL: name: clause_implicit_def
; GCN-CLAUSE: liveins: $vgpr0, $sgpr32
; GCN-CLAUSE-NEXT: {{ $}}
- ; GCN-CLAUSE-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
- ; GCN-CLAUSE-NEXT: $vgpr2 = IMPLICIT_DEF
- ; GCN-CLAUSE-NEXT: $vgpr3 = IMPLICIT_DEF
- ; GCN-CLAUSE-NEXT: SCRATCH_STORE_DWORDX2_SADDR $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+ ; GCN-CLAUSE-NEXT: BUNDLE implicit-def $vgpr2, implicit-def $vgpr2_lo16, implicit-def $vgpr2_hi16, implicit-def $vgpr3, implicit-def $vgpr3_lo16, implicit-def $vgpr3_hi16, implicit $vgpr0, implicit $sgpr32, implicit $exec, implicit $flat_scr {
+ ; GCN-CLAUSE-NEXT: S_CLAUSE 1
+ ; GCN-CLAUSE-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
+ ; GCN-CLAUSE-NEXT: $vgpr2 = IMPLICIT_DEF
+ ; GCN-CLAUSE-NEXT: $vgpr3 = IMPLICIT_DEF
+ ; GCN-CLAUSE-NEXT: SCRATCH_STORE_DWORDX2_SADDR internal $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+ ; GCN-CLAUSE-NEXT: }
;
; GCN-BUNDLE-LABEL: name: clause_implicit_def
; GCN-BUNDLE: liveins: $vgpr0, $sgpr32
; GCN-BUNDLE-NEXT: {{ $}}
- ; GCN-BUNDLE-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
- ; GCN-BUNDLE-NEXT: $vgpr2 = IMPLICIT_DEF
- ; GCN-BUNDLE-NEXT: $vgpr3 = IMPLICIT_DEF
- ; GCN-BUNDLE-NEXT: SCRATCH_STORE_DWORDX2_SADDR $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+ ; GCN-BUNDLE-NEXT: BUNDLE implicit-def $vgpr2, implicit-def $vgpr2_lo16, implicit-def $vgpr2_hi16, implicit-def $vgpr3, implicit-def $vgpr3_lo16, implicit-def $vgpr3_hi16, implicit $vgpr0, implicit $sgpr32, implicit $exec, implicit $flat_scr {
+ ; GCN-BUNDLE-NEXT: SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
+ ; GCN-BUNDLE-NEXT: $vgpr2 = IMPLICIT_DEF
+ ; GCN-BUNDLE-NEXT: $vgpr3 = IMPLICIT_DEF
+ ; GCN-BUNDLE-NEXT: SCRATCH_STORE_DWORDX2_SADDR internal $vgpr2_vgpr3, $sgpr32, 16, 0, implicit $exec, implicit $flat_scr
+ ; GCN-BUNDLE-NEXT: }
SCRATCH_STORE_DWORD_SADDR $vgpr0, $sgpr32, 8, 0, implicit $exec, implicit $flat_scr
$vgpr2 = IMPLICIT_DEF
$vgpr3 = IMPLICIT_DEF
More information about the llvm-commits
mailing list