This is an archive of the discontinued LLVM Phabricator instance.

Differential D156125

[AMDGPU] Fix llvm.amdgcn.wave.reduce.umax/umin MIR tests
ClosedPublic

Authored by pravinjagtap on Jul 24 2023, 7:15 AM.

Details

Reviewers
arsenm
foad
ruiling
Pierre-vh
yassingh
b-sumner
cdevadas
Group Reviewers
Restricted Project
Commits
rGd163b76ce348: [AMDGPU] Fix llvm.amdgcn.wave.reduce.umax/umin MIR tests
Summary

Fixes the MIR tests reported in https://lab.llvm.org/buildbot/#/builders/16/builds/51955

Diff Detail

Event Timeline

pravinjagtap created this revision.Jul 24 2023, 7:15 AM
Herald added a project: Restricted Project. · View Herald TranscriptJul 24 2023, 7:15 AM
Herald added subscribers: kerbowa, tpr, dstuttard and 3 others. · View Herald Transcript
pravinjagtap requested review of this revision.Jul 24 2023, 7:15 AM
Herald added a project: Restricted Project. · View Herald TranscriptJul 24 2023, 7:15 AM
Herald added subscribers: llvm-commits, wdng. · View Herald Transcript
arsenm accepted this revision.Jul 24 2023, 7:16 AM
This revision is now accepted and ready to land.Jul 24 2023, 7:16 AM
pravinjagtap edited the summary of this revision. (Show Details)Jul 24 2023, 7:18 AM
pravinjagtap added reviewers: foad, ruiling, Pierre-vh, yassingh, b-sumner, cdevadas, Restricted Project.
Herald added a subscriber: StephenFan. · View Herald TranscriptJul 24 2023, 7:18 AM
This revision was landed with ongoing or failed builds.Jul 24 2023, 7:20 AM
Closed by commit rGd163b76ce348: [AMDGPU] Fix llvm.amdgcn.wave.reduce.umax/umin MIR tests (authored by pravinjagtap). · Explain Why
This revision was automatically updated to reflect the committed changes.
pravinjagtap added a commit: rGd163b76ce348: [AMDGPU] Fix llvm.amdgcn.wave.reduce.umax/umin MIR tests.
Harbormaster completed remote builds in B247668: Diff 543533.Jul 24 2023, 10:59 AM

Revision Contents

PathSize
llvm/
test/
CodeGen/
AMDGPU/
68 lines
68 lines

Diff 543538

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wave.reduce.umax.mir

# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3
# RUN: llc -march=amdgcn -run-pass=finalize-isel %s -o - | FileCheck -check-prefix=GCN %s # RUN: llc -march=amdgcn -run-pass=finalize-isel -verify-machineinstrs %s -o - | FileCheck -check-prefix=GCN %s
--- ---
name: uniform_value name: uniform_value
tracksRegLiveness: true tracksRegLiveness: true
machineFunctionInfo: machineFunctionInfo:
isEntryFunction: true isEntryFunction: true
body: | body: |
bb.0.entry: bb.0.entry:
Show All 20 Lines bb.0.entry:
S_ENDPGM 0 S_ENDPGM 0
... ...
--- ---
name: divergent_value name: divergent_value
machineFunctionInfo: machineFunctionInfo:
isEntryFunction: true isEntryFunction: true
tracksRegLiveness: true
body: | body: |
bb.0.entry:
liveins: $vgpr0, $sgpr0_sgpr1
; GCN-LABEL: name: divergent_value ; GCN-LABEL: name: divergent_value
; GCN: successors: %bb.1(0x80000000) ; GCN: bb.0.entry:
; GCN-NEXT: successors: %bb.2(0x80000000)
; GCN-NEXT: liveins: $vgpr0, $sgpr0_sgpr1 ; GCN-NEXT: liveins: $vgpr0, $sgpr0_sgpr1
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: [[COPY:%[0-9]+]]:sgpr_64(p4) = COPY $sgpr0_sgpr1 ; GCN-NEXT: [[COPY:%[0-9]+]]:sgpr_64(p4) = COPY $sgpr0_sgpr1
; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GCN-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]](p4), 36, 0 ; GCN-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]](p4), 36, 0
; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec ; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
; GCN-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec ; GCN-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
; GCN-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 0 ; GCN-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 0
; GCN-NEXT: S_BRANCH %bb.1 ; GCN-NEXT: S_BRANCH %bb.2
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: .1: ; GCN-NEXT: bb.2:
; GCN-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) ; GCN-NEXT: successors: %bb.2(0x40000000), %bb.3(0x40000000)
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: [[PHI:%[0-9]+]]:sgpr_32 = PHI [[S_MOV_B32_]], %bb.0, %4, %bb.1 ; GCN-NEXT: [[PHI:%[0-9]+]]:sgpr_32 = PHI [[S_MOV_B32_]], %bb.0, %4, %bb.2
; GCN-NEXT: [[PHI1:%[0-9]+]]:sreg_64_xexec = PHI [[S_MOV_B64_]], %bb.0, %10, %bb.1 ; GCN-NEXT: [[PHI1:%[0-9]+]]:sreg_64_xexec = PHI [[S_MOV_B64_]], %bb.0, %11, %bb.2
; GCN-NEXT: [[S_FF1_I32_B64_:%[0-9]+]]:sgpr_32 = S_FF1_I32_B64 [[PHI1]] ; GCN-NEXT: [[S_FF1_I32_B64_:%[0-9]+]]:sgpr_32 = S_FF1_I32_B64 [[PHI1]]
; GCN-NEXT: [[V_READLANE_B32_:%[0-9]+]]:sgpr_32 = V_READLANE_B32 [[COPY1]], [[S_FF1_I32_B64_]] ; GCN-NEXT: [[V_READLANE_B32_:%[0-9]+]]:sgpr_32 = V_READLANE_B32 [[COPY1]], [[S_FF1_I32_B64_]]
; GCN-NEXT: [[S_MAX_U32_:%[0-9]+]]:sgpr_32 = S_MAX_U32 [[PHI]], [[V_READLANE_B32_]], implicit-def $scc ; GCN-NEXT: [[S_MAX_U32_:%[0-9]+]]:sgpr_32 = S_MAX_U32 [[PHI]], [[V_READLANE_B32_]], implicit-def $scc
; GCN-NEXT: [[S_BITSET0_B64_:%[0-9]+]]:sreg_64_xexec = S_BITSET0_B64 [[S_FF1_I32_B64_]], [[PHI1]] ; GCN-NEXT: [[S_BITSET0_B64_:%[0-9]+]]:sreg_64_xexec = S_BITSET0_B64 [[S_FF1_I32_B64_]], [[PHI1]]
; GCN-NEXT: S_CMP_LG_U64 [[S_BITSET0_B64_]], 0, implicit-def $scc ; GCN-NEXT: S_CMP_LG_U64 [[S_BITSET0_B64_]], 0, implicit-def $scc
; GCN-NEXT: S_CBRANCH_SCC1 %bb.1, implicit $scc ; GCN-NEXT: S_CBRANCH_SCC1 %bb.2, implicit $scc
; GCN-NEXT: {{ $}}
; GCN-NEXT: bb.3:
; GCN-NEXT: successors: %bb.1(0x80000000)
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: .2:
; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[S_MAX_U32_]] ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[S_MAX_U32_]]
; GCN-NEXT: GLOBAL_STORE_DWORD_SADDR killed [[V_MOV_B32_e32_]], killed [[COPY2]], killed [[S_LOAD_DWORDX2_IMM]], 0, 0, implicit $exec ; GCN-NEXT: GLOBAL_STORE_DWORD_SADDR killed [[V_MOV_B32_e32_]], killed [[COPY2]], killed [[S_LOAD_DWORDX2_IMM]], 0, 0, implicit $exec
; GCN-NEXT: {{ $}}
; GCN-NEXT: bb.1:
; GCN-NEXT: [[PHI2:%[0-9]+]]:vgpr_32 = PHI [[COPY1]], %bb.3
; GCN-NEXT: S_ENDPGM 0 ; GCN-NEXT: S_ENDPGM 0
bb.0.entry:
liveins: $vgpr0, $sgpr0_sgpr1
%1:sgpr_64(p4) = COPY $sgpr0_sgpr1 %1:sgpr_64(p4) = COPY $sgpr0_sgpr1
%0:vgpr_32 = COPY $vgpr0 %0:vgpr_32 = COPY $vgpr0
%4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %1(p4), 36, 0 %4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %1(p4), 36, 0
%5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
%6:sgpr_32 = WAVE_REDUCE_UMAX_PSEUDO_U32 %0, 1, implicit $exec %6:sgpr_32 = WAVE_REDUCE_UMAX_PSEUDO_U32 %0, 1, implicit $exec
%7:vgpr_32 = COPY %6 %7:vgpr_32 = COPY %6
GLOBAL_STORE_DWORD_SADDR killed %5, killed %7, killed %4, 0, 0, implicit $exec GLOBAL_STORE_DWORD_SADDR killed %5, killed %7, killed %4, 0, 0, implicit $exec
bb.1:
%8:vgpr_32 = PHI %0, %bb.0
S_ENDPGM 0 S_ENDPGM 0
... ...

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wave.reduce.umin.mir

# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3
# RUN: llc -march=amdgcn -run-pass=finalize-isel %s -o - | FileCheck -check-prefix=GCN %s # RUN: llc -march=amdgcn -run-pass=finalize-isel -verify-machineinstrs %s -o - | FileCheck -check-prefix=GCN %s
--- ---
name: uniform_value name: uniform_value
tracksRegLiveness: true tracksRegLiveness: true
machineFunctionInfo: machineFunctionInfo:
isEntryFunction: true isEntryFunction: true
body: | body: |
bb.0.entry: bb.0.entry:
Show All 20 Lines bb.0.entry:
S_ENDPGM 0 S_ENDPGM 0
... ...
--- ---
name: divergent_value name: divergent_value
machineFunctionInfo: machineFunctionInfo:
isEntryFunction: true isEntryFunction: true
tracksRegLiveness: true
body: | body: |
bb.0.entry:
liveins: $vgpr0, $sgpr0_sgpr1
; GCN-LABEL: name: divergent_value ; GCN-LABEL: name: divergent_value
; GCN: successors: %bb.1(0x80000000) ; GCN: bb.0.entry:
; GCN-NEXT: successors: %bb.2(0x80000000)
; GCN-NEXT: liveins: $vgpr0, $sgpr0_sgpr1 ; GCN-NEXT: liveins: $vgpr0, $sgpr0_sgpr1
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: [[COPY:%[0-9]+]]:sgpr_64(p4) = COPY $sgpr0_sgpr1 ; GCN-NEXT: [[COPY:%[0-9]+]]:sgpr_64(p4) = COPY $sgpr0_sgpr1
; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GCN-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]](p4), 36, 0 ; GCN-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]](p4), 36, 0
; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec ; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
; GCN-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec ; GCN-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
; GCN-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 4294967295 ; GCN-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 4294967295
; GCN-NEXT: S_BRANCH %bb.1 ; GCN-NEXT: S_BRANCH %bb.2
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: .1: ; GCN-NEXT: bb.2:
; GCN-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) ; GCN-NEXT: successors: %bb.2(0x40000000), %bb.3(0x40000000)
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: [[PHI:%[0-9]+]]:sgpr_32 = PHI [[S_MOV_B32_]], %bb.0, %4, %bb.1 ; GCN-NEXT: [[PHI:%[0-9]+]]:sgpr_32 = PHI [[S_MOV_B32_]], %bb.0, %4, %bb.2
; GCN-NEXT: [[PHI1:%[0-9]+]]:sreg_64_xexec = PHI [[S_MOV_B64_]], %bb.0, %10, %bb.1 ; GCN-NEXT: [[PHI1:%[0-9]+]]:sreg_64_xexec = PHI [[S_MOV_B64_]], %bb.0, %11, %bb.2
; GCN-NEXT: [[S_FF1_I32_B64_:%[0-9]+]]:sgpr_32 = S_FF1_I32_B64 [[PHI1]] ; GCN-NEXT: [[S_FF1_I32_B64_:%[0-9]+]]:sgpr_32 = S_FF1_I32_B64 [[PHI1]]
; GCN-NEXT: [[V_READLANE_B32_:%[0-9]+]]:sgpr_32 = V_READLANE_B32 [[COPY1]], [[S_FF1_I32_B64_]] ; GCN-NEXT: [[V_READLANE_B32_:%[0-9]+]]:sgpr_32 = V_READLANE_B32 [[COPY1]], [[S_FF1_I32_B64_]]
; GCN-NEXT: [[S_MIN_U32_:%[0-9]+]]:sgpr_32 = S_MIN_U32 [[PHI]], [[V_READLANE_B32_]], implicit-def $scc ; GCN-NEXT: [[S_MIN_U32_:%[0-9]+]]:sgpr_32 = S_MIN_U32 [[PHI]], [[V_READLANE_B32_]], implicit-def $scc
; GCN-NEXT: [[S_BITSET0_B64_:%[0-9]+]]:sreg_64_xexec = S_BITSET0_B64 [[S_FF1_I32_B64_]], [[PHI1]] ; GCN-NEXT: [[S_BITSET0_B64_:%[0-9]+]]:sreg_64_xexec = S_BITSET0_B64 [[S_FF1_I32_B64_]], [[PHI1]]
; GCN-NEXT: S_CMP_LG_U64 [[S_BITSET0_B64_]], 0, implicit-def $scc ; GCN-NEXT: S_CMP_LG_U64 [[S_BITSET0_B64_]], 0, implicit-def $scc
; GCN-NEXT: S_CBRANCH_SCC1 %bb.1, implicit $scc ; GCN-NEXT: S_CBRANCH_SCC1 %bb.2, implicit $scc
; GCN-NEXT: {{ $}}
; GCN-NEXT: bb.3:
; GCN-NEXT: successors: %bb.1(0x80000000)
; GCN-NEXT: {{ $}} ; GCN-NEXT: {{ $}}
; GCN-NEXT: .2:
; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[S_MIN_U32_]] ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[S_MIN_U32_]]
; GCN-NEXT: GLOBAL_STORE_DWORD_SADDR killed [[V_MOV_B32_e32_]], killed [[COPY2]], killed [[S_LOAD_DWORDX2_IMM]], 0, 0, implicit $exec ; GCN-NEXT: GLOBAL_STORE_DWORD_SADDR killed [[V_MOV_B32_e32_]], killed [[COPY2]], killed [[S_LOAD_DWORDX2_IMM]], 0, 0, implicit $exec
; GCN-NEXT: {{ $}}
; GCN-NEXT: bb.1:
; GCN-NEXT: [[PHI2:%[0-9]+]]:vgpr_32 = PHI [[COPY1]], %bb.3
; GCN-NEXT: S_ENDPGM 0 ; GCN-NEXT: S_ENDPGM 0
bb.0.entry:
liveins: $vgpr0, $sgpr0_sgpr1
%1:sgpr_64(p4) = COPY $sgpr0_sgpr1 %1:sgpr_64(p4) = COPY $sgpr0_sgpr1
%0:vgpr_32 = COPY $vgpr0 %0:vgpr_32 = COPY $vgpr0
%4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %1(p4), 36, 0 %4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %1(p4), 36, 0
%5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec %5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
%6:sgpr_32 = WAVE_REDUCE_UMIN_PSEUDO_U32 %0, 1, implicit $exec %6:sgpr_32 = WAVE_REDUCE_UMIN_PSEUDO_U32 %0, 1, implicit $exec
%7:vgpr_32 = COPY %6 %7:vgpr_32 = COPY %6
GLOBAL_STORE_DWORD_SADDR killed %5, killed %7, killed %4, 0, 0, implicit $exec GLOBAL_STORE_DWORD_SADDR killed %5, killed %7, killed %4, 0, 0, implicit $exec
bb.1:
%8:vgpr_32 = PHI %0, %bb.0
S_ENDPGM 0 S_ENDPGM 0
... ...