Skip to content

[AArch64][GlobalISel] Add support for widening vector store elements to s8. #121170

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

Merged
merged 2 commits into from
Jan 6, 2025
Merged

[AArch64][GlobalISel] Add support for widening vector store elements to s8. #121170

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4b9f425
[𝘀𝗽𝗿] initial version
aemerson Dec 27, 2024
d69e551
Add assert against poiters.
aemerson Dec 27, 2024
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
15 changes: 13 additions & 2 deletions llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3022,8 +3022,19 @@ LegalizerHelper::widenScalar(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) {
return UnableToLegalize;

LLT Ty = MRI.getType(MI.getOperand(0).getReg());
if (!Ty.isScalar())
return UnableToLegalize;
assert(!Ty.isPointerOrPointerVector() && "Can't widen type");
if (!Ty.isScalar()) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is going to break on the pointer case?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I can put an assert, but it shouldn't be valid to widenScalarOrElt on a pointer vector?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

True

// We need to widen the vector element type.
Observer.changingInstr(MI);
widenScalarSrc(MI, WideTy, 0, TargetOpcode::G_ANYEXT);
// We also need to adjust the MMO to turn this into a truncating store.
MachineMemOperand &MMO = **MI.memoperands_begin();
MachineFunction &MF = MIRBuilder.getMF();
auto *NewMMO = MF.getMachineMemOperand(&MMO, MMO.getPointerInfo(), Ty);
MI.setMemRefs(MF, {NewMMO});
Observer.changedInstr(MI);
return Legalized;
}

Observer.changingInstr(MI);

Expand Down
1 change: 1 addition & 0 deletions llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -454,6 +454,7 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
{nxv2s64, p0, nxv2s64, 8},
})
.clampScalar(0, s8, s64)
.minScalarOrElt(0, s8)
.lowerIf([=](const LegalityQuery &Query) {
return Query.Types[0].isScalar() &&
Query.Types[0] != Query.MMODescrs[0].MemoryTy;
Expand Down
32 changes: 32 additions & 0 deletions llvm/test/CodeGen/AArch64/GlobalISel/legalize-store-vector-bools.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
# RUN: llc -O0 -mtriple=aarch64 -run-pass=legalizer -global-isel-abort=2 %s -o - | FileCheck %s
# This test currently is expected to fall back after reaching truncstore of <8 x s8> as <8 x s1>.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a fragile way of testing this fallback case (testing fallback in a single pass MIR test is also odd). Should check the fallback error message, unless you're just going to fix this in the next commit anyway

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a fragile way of testing this fallback case (testing fallback in a single pass MIR test is also odd). Should check the fallback error message, unless you're just going to fix this in the next commit anyway

Yeah this is going to be immediately fixed in the follow ups.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also, I'm not testing that it falls back. I'm trying to test the intermediate legalizer output but a fallback prevents the full legalizer pass running successfully. This is a workaround so we can still check the widening was done.

---
name: store_8xs1
tracksRegLiveness: true
body: |
bb.1:
liveins: $q0, $q1, $x0
; CHECK-LABEL: name: store_8xs1
; CHECK: liveins: $q0, $q1, $x0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1
; CHECK-NEXT: %ptr:_(p0) = COPY $x0
; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s32>) = G_CONCAT_VECTORS [[COPY]](<4 x s32>), [[COPY1]](<4 x s32>)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(<8 x s1>) = G_ICMP intpred(slt), [[CONCAT_VECTORS]](<8 x s32>), [[BUILD_VECTOR]]
; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(<8 x s8>) = G_ANYEXT [[ICMP]](<8 x s1>)
; CHECK-NEXT: G_STORE [[ANYEXT]](<8 x s8>), %ptr(p0) :: (store (<8 x s1>))
; CHECK-NEXT: RET_ReallyLR
%1:_(<4 x s32>) = COPY $q0
%2:_(<4 x s32>) = COPY $q1
%ptr:_(p0) = COPY $x0
%0:_(<8 x s32>) = G_CONCAT_VECTORS %1(<4 x s32>), %2(<4 x s32>)
%4:_(s32) = G_CONSTANT i32 0
%3:_(<8 x s32>) = G_BUILD_VECTOR %4(s32), %4(s32), %4(s32), %4(s32), %4(s32), %4(s32), %4(s32), %4(s32)
%5:_(<8 x s1>) = G_ICMP intpred(slt), %0(<8 x s32>), %3
G_STORE %5(<8 x s1>), %ptr(p0) :: (store (<8 x s1>))
RET_ReallyLR
...
Loading