Skip to content

[InstCombine][foldPHIArgGEPIntoPHI] Early return for const vector index for gep inst. #138661

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 4 commits into from
May 8, 2025
Merged

[InstCombine][foldPHIArgGEPIntoPHI] Early return for const vector index for gep inst. #138661

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5569ace
[InstCombine][foldPHIArgGEPIntoPHI] Early return for const vector ind…
haonanya1 May 6, 2025
8860620
Apply suggestions
haonanya1 May 7, 2025
abb6564
Update lit.
haonanya1 May 7, 2025
06c7fe2
Update lit.
haonanya1 May 8, 2025
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
4 changes: 2 additions & 2 deletions llvm/lib/Transforms/InstCombine/InstCombinePHI.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -574,8 +574,8 @@ Instruction *InstCombinerImpl::foldPHIArgGEPIntoPHI(PHINode &PN) {
// substantially cheaper to compute for the constants, so making it a
// variable index could pessimize the path. This also handles the case
// for struct indices, which must always be constant.
if (isa<ConstantInt>(FirstInst->getOperand(Op)) ||
isa<ConstantInt>(GEP->getOperand(Op)))
if (isa<Constant>(FirstInst->getOperand(Op)) ||
isa<Constant>(GEP->getOperand(Op)))
return nullptr;

if (FirstInst->getOperand(Op)->getType() !=
Expand Down
39 changes: 39 additions & 0 deletions llvm/test/Transforms/InstCombine/fold-phi-arg-gep-to-phi-negative.ll
View file
Open in desktop
Copy link
Contributor

Choose a reason for hiding this comment

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

This ended up a bit too reduced now and no longer crashes without the change. This variant works:

%foo = type { i16, i16, i16 }

define <16 x ptr> @test(i1 %tobool) {
entry:
  %lane.0 = alloca %foo, align 16
  %lane.15 = insertelement <16 x ptr> poison, ptr %lane.0, i64 0
  %mm_vectorGEP = getelementptr inbounds %foo, <16 x ptr> %lane.15, <16 x i64> zeroinitializer, <16 x i32> splat (i32 1)
  %mm_vectorGEP2 = getelementptr inbounds %foo, <16 x ptr> %lane.15, <16 x i64> zeroinitializer, <16 x i32> splat (i32 2)
  br i1 %tobool, label %f1, label %f0

f0:
  br label %merge

f1:
  br label %merge

merge:
  %vec.phi = phi <16 x ptr> [ %mm_vectorGEP, %f0], [ %mm_vectorGEP2, %f1 ]
  ret <16 x ptr> %vec.phi
}

(This makes sure the second GEP is not optimized away.)

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Thanks for your patience, updated.

Original file line number Diff line number Diff line change
@@ -0,0 +1,39 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt < %s -passes=instcombine -S | FileCheck %s

%foo = type { i16, i16, i16 }

define <16 x ptr> @test(i1 %tobool) {
; CHECK-LABEL: define <16 x ptr> @test(
; CHECK-SAME: i1 [[TOBOOL:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: [[LANE_0:%.*]] = alloca [[FOO:%.*]], align 16
; CHECK-NEXT: [[LANE_15:%.*]] = insertelement <16 x ptr> poison, ptr [[LANE_0]], i64 0
; CHECK-NEXT: br i1 [[TOBOOL]], label %[[F1:.*]], label %[[F0:.*]]
; CHECK: [[F0]]:
; CHECK-NEXT: [[MM_VECTORGEP:%.*]] = getelementptr inbounds [[FOO]], <16 x ptr> [[LANE_15]], <16 x i64> zeroinitializer, <16 x i32> splat (i32 1)
; CHECK-NEXT: br label %[[MERGE:.*]]
; CHECK: [[F1]]:
; CHECK-NEXT: [[MM_VECTORGEP2:%.*]] = getelementptr inbounds [[FOO]], <16 x ptr> [[LANE_15]], <16 x i64> zeroinitializer, <16 x i32> splat (i32 2)
; CHECK-NEXT: br label %[[MERGE]]
; CHECK: [[MERGE]]:
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <16 x ptr> [ [[MM_VECTORGEP]], %[[F0]] ], [ [[MM_VECTORGEP2]], %[[F1]] ]
; CHECK-NEXT: ret <16 x ptr> [[VEC_PHI]]
;
entry:
%lane.0 = alloca %foo, align 16
%lane.15 = insertelement <16 x ptr> poison, ptr %lane.0, i64 0
%mm_vectorGEP = getelementptr inbounds %foo, <16 x ptr> %lane.15, <16 x i64> zeroinitializer, <16 x i32> splat (i32 1)
%mm_vectorGEP2 = getelementptr inbounds %foo, <16 x ptr> %lane.15, <16 x i64> zeroinitializer, <16 x i32> splat (i32 2)
br i1 %tobool, label %f1, label %f0

f0:
br label %merge

f1:
br label %merge

merge:
%vec.phi = phi <16 x ptr> [ %mm_vectorGEP, %f0], [ %mm_vectorGEP2, %f1 ]
ret <16 x ptr> %vec.phi
}