maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

AMDGPU/GFX10: Apply the VMEM-to-scalar-write hazard also to writes to EXEC 

Summary: Change-Id: I854fbf7d48e937bef9f8f3f5d0c8aeb970652630

Reviewers: rampitec, mareko

Subscribers: arsenm, kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D64807

Change-Id: I4405b3a7f84186acea5a78d291bff71056e745fc
llvm-svn: 366314
This commit is contained in:
Nicolai Haehnle 2019-07-17 11:22:57 +00:00
parent a256b8b7d7
commit 8b7041a5c6
4 changed files with 18 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp
View File

@ -920,7 +920,7 @@ bool GCNHazardRecognizer::fixVMEMtoScalarWriteHazards(MachineInstr *MI) {
for (const MachineOperand &Def : MI->defs()) {
MachineOperand *Op = I->findRegisterUseOperand(Def.getReg(), false, TRI);
if (!Op || (Op->isImplicit() && Op->getReg() == AMDGPU::EXEC))
if (!Op)
continue;
return true;
}

1
llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.end.cf.i32.ll
View File

@ -14,6 +14,7 @@ define amdgpu_kernel void @test_wave32(i32 %arg0, [8 x i32], i32 %saved) {
; GCN-NEXT: v_mov_b32_e32 v0, 0
; GCN-NEXT: global_store_dword v[0:1], v0, off
; GCN-NEXT: BB0_2: ; %bb
; GCN-NEXT: v_nop
; GCN-NEXT: s_or_b32 exec_lo, exec_lo, s0
; GCN-NEXT: v_mov_b32_e32 v0, 0
; GCN-NEXT: global_store_dword v[0:1], v0, off

14
llvm/test/CodeGen/AMDGPU/vmem-to-salu-hazard.mir
View File

@ -92,6 +92,7 @@ body: |
...
# GCN-LABEL: name: vmem_write_exec_impread
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN: V_NOP
# GCN-NEXT: S_MOV_B64
---
name: vmem_write_exec_impread
@ -208,3 +209,16 @@ body: |
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, implicit $exec
S_BRANCH %bb.0
...
# GCN-LABEL: name: ds_write_exec
# GCN: DS_WRITE_B32_gfx9
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: ds_write_exec
body: |
bb.0:
$vgpr0 = IMPLICIT_DEF
$vgpr1 = IMPLICIT_DEF
DS_WRITE_B32_gfx9 $vgpr0, $vgpr1, 0, 0, implicit $exec
$exec_lo = S_MOV_B32 -1
...

2
llvm/test/CodeGen/AMDGPU/wave32.ll
View File

@ -1073,6 +1073,7 @@ declare void @external_void_func_void() #1
; GFX1064-NEXT: s_or_saveexec_b64 [[COPY_EXEC0:s\[[0-9]+:[0-9]+\]]], -1{{$}}
; GFX1032-NEXT: s_or_saveexec_b32 [[COPY_EXEC0:s[0-9]]], -1{{$}}
; GCN-NEXT: buffer_store_dword v32, off, s[0:3], s32 ; 4-byte Folded Spill
; GCN-NEXT: v_nop
; GFX1064-NEXT: s_mov_b64 exec, [[COPY_EXEC0]]
; GFX1032-NEXT: s_mov_b32 exec_lo, [[COPY_EXEC0]]
@ -1095,6 +1096,7 @@ declare void @external_void_func_void() #1
; GFX1064: s_or_saveexec_b64 [[COPY_EXEC1:s\[[0-9]+:[0-9]+\]]], -1{{$}}
; GFX1032: s_or_saveexec_b32 [[COPY_EXEC1:s[0-9]]], -1{{$}}
; GCN-NEXT: buffer_load_dword v32, off, s[0:3], s32 ; 4-byte Folded Reload
; GCN-NEXT: v_nop
; GFX1064-NEXT: s_mov_b64 exec, [[COPY_EXEC1]]
; GFX1032-NEXT: s_mov_b32 exec_lo, [[COPY_EXEC1]]
; GCN-NEXT: s_waitcnt vmcnt(0)