Skip to content

[Legalizer] Cache extracted element when lowering G_SHUFFLE_VECTOR. #163893

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[Legalizer] Cache extracted element when lowering G_SHUFFLE_VECTOR. #163893

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 9 additions & 3 deletions llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9008,6 +9008,8 @@ LegalizerHelper::lowerShuffleVector(MachineInstr &MI) {
SmallVector<Register, 32> BuildVec;
LLT EltTy = DstTy.getScalarType();

DenseMap<unsigned, Register> CachedExtract;

for (int Idx : Mask) {
if (Idx < 0) {
if (!Undef.isValid())
Expand All @@ -9022,9 +9024,13 @@ LegalizerHelper::lowerShuffleVector(MachineInstr &MI) {
int NumElts = Src0Ty.getNumElements();
Register SrcVec = Idx < NumElts ? Src0Reg : Src1Reg;
int ExtractIdx = Idx < NumElts ? Idx : Idx - NumElts;
auto IdxK = MIRBuilder.buildConstant(IdxTy, ExtractIdx);
auto Extract = MIRBuilder.buildExtractVectorElement(EltTy, SrcVec, IdxK);
BuildVec.push_back(Extract.getReg(0));
auto [It, Inserted] = CachedExtract.try_emplace(Idx);
if (Inserted) {
auto IdxK = MIRBuilder.buildConstant(IdxTy, ExtractIdx);
It->second =
MIRBuilder.buildExtractVectorElement(EltTy, SrcVec, IdxK).getReg(0);
}
BuildVec.push_back(It->second);
}
}

Expand Down
31 changes: 28 additions & 3 deletions llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shuffle-vector.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,9 +136,7 @@ body: |
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3
; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
; CHECK-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY3]](s32)
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY2]](s32)
; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(<2 x s32>) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
Expand Down Expand Up @@ -374,3 +372,30 @@ body: |
$vgpr0_vgpr1 = COPY %4

...

---
name: shufflevector_cache_element
tracksRegLiveness: true

body: |
bb.0:
liveins: $vgpr0_vgpr1, $vgpr2_vgpr3

; CHECK-LABEL: name: shufflevector_cache_element
; CHECK: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY2]](s32), [[COPY2]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY3]](s32), [[COPY3]](s32), [[COPY3]](s32)
; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[BUILD_VECTOR]](<8 x s32>), [[COPY1]](s32)
; CHECK-NEXT: $vgpr0 = COPY [[EVEC]](s32)
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(<2 x s32>) = G_BUILD_VECTOR %0, %1
%3:_(<8 x s32>) = G_SHUFFLE_VECTOR %2, %2, shufflemask(0, 0, 0, 0, 3, 3, 3, 3)
%4:_(s32) = G_EXTRACT_VECTOR_ELT %3(<8 x s32>), %1(s32)
$vgpr0 = COPY %4

...
68 changes: 20 additions & 48 deletions llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shuffle-vector.s16.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -237,16 +237,13 @@ body: |
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C2]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
;
; GFX9-LABEL: name: shufflevector_v2s16_v2s16_0_0
; GFX9: liveins: $vgpr0, $vgpr1
Expand All @@ -257,10 +254,7 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
Expand All @@ -285,12 +279,10 @@ body: |
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[LSHR]], [[SHL]]
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
;
; GFX9-LABEL: name: shufflevector_v2s16_v2s16_1_1
; GFX9: liveins: $vgpr0, $vgpr1
Expand All @@ -301,10 +293,7 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
Expand All @@ -329,16 +318,13 @@ body: |
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C2]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
;
; GFX9-LABEL: name: shufflevector_v2s16_v2s16_2_2
; GFX9: liveins: $vgpr0, $vgpr1
Expand All @@ -349,10 +335,7 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
Expand Down Expand Up @@ -641,12 +624,10 @@ body: |
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[LSHR]], [[SHL]]
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
;
; GFX9-LABEL: name: shufflevector_v2s16_v2s16_3_3
; GFX9: liveins: $vgpr0, $vgpr1
Expand All @@ -657,10 +638,7 @@ body: |
; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
Expand Down Expand Up @@ -733,16 +711,13 @@ body: |
; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C2]](s32)
; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
; GFX8-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
;
; GFX9-LABEL: name: shufflevector_v2s16_v2s16_2_0
; GFX9: liveins: $vgpr0, $vgpr1
Expand All @@ -753,10 +728,7 @@ body: |
; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16)
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC]](s16)
; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR]](<2 x s16>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
Expand Down