Skip to content

[SLP][REVEC] Ignore UserTreeIndex if it is empty. #131993

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
Mar 20, 2025
Merged

[SLP][REVEC] Ignore UserTreeIndex if it is empty. #131993

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
e92f2ee
[SLP][REVEC] Pre-commit test.
HanKuanChen Mar 19, 2025
23ad559
[SLP][REVEC] Ignore UserTreeIndex if it is empty.
HanKuanChen Mar 19, 2025
bd6d98c
remove assert
HanKuanChen Mar 19, 2025
11dde44
apply comment
HanKuanChen Mar 19, 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
11 changes: 4 additions & 7 deletions llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6449,14 +6449,11 @@ void BoUpSLP::reorderTopToBottom() {
assert(SLPReVec && "Only supported by REVEC.");
// ShuffleVectorInst does not do reorderOperands (and it should not
// because ShuffleVectorInst supports only a limited set of
// patterns). Only do reorderNodeWithReuses if all of the users are
// not ShuffleVectorInst.
if (isa<ShuffleVectorInst>(TE->UserTreeIndex.UserTE->getMainOp()))
// patterns). Only do reorderNodeWithReuses if the user is not
// ShuffleVectorInst.
if (TE->UserTreeIndex && TE->UserTreeIndex.UserTE->hasState() &&
isa<ShuffleVectorInst>(TE->UserTreeIndex.UserTE->getMainOp()))
Copy link
Member

Choose a reason for hiding this comment

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

Some of the user may not have state, need to add a check for hasState() before asking for getMainOp

Copy link
Contributor Author

Choose a reason for hiding this comment

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

How? If a TreeEntry has a user, that means the user must have MainOp inside InstructionsState. Otherwise, buildTree_rec cannot work.

Copy link
Member

Choose a reason for hiding this comment

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

Some entries might have Gather parents

continue;
assert((!TE->UserTreeIndex ||
!isa<ShuffleVectorInst>(
TE->UserTreeIndex.UserTE->getMainOp())) &&
"Does not know how to reorder.");
}
// Update ordering of the operands with the smaller VF than the given
// one.
Expand Down
35 changes: 35 additions & 0 deletions llvm/test/Transforms/SLPVectorizer/X86/revec-reduced-value-replace-extractelement.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S --passes=slp-vectorizer -slp-threshold=-99999 < %s -mtriple=x86_64-unknown-linux-gnu -slp-revec | FileCheck %s

define void @test() {
; CHECK-LABEL: define void @test() {
; CHECK-NEXT: [[BB:.*]]:
; CHECK-NEXT: [[TRUNC:%.*]] = trunc i64 0 to i32
; CHECK-NEXT: br label %[[BB1:.*]]
; CHECK: [[BB1]]:
; CHECK-NEXT: [[TMP0:%.*]] = phi <2 x i32> [ zeroinitializer, %[[BB]] ], [ [[TMP4:%.*]], %[[BB1]] ]
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x i32> [[TMP0]], i32 1
; CHECK-NEXT: [[TMP2:%.*]] = call i8 @llvm.vector.reduce.mul.v4i8(<4 x i8> zeroinitializer)
; CHECK-NEXT: [[TMP3:%.*]] = zext i8 [[TMP2]] to i32
; CHECK-NEXT: [[OP_RDX:%.*]] = mul i32 [[TMP3]], [[TMP1]]
; CHECK-NEXT: [[OP_RDX1:%.*]] = mul i32 [[OP_RDX]], [[TRUNC]]
; CHECK-NEXT: [[TMP4]] = insertelement <2 x i32> <i32 0, i32 poison>, i32 [[OP_RDX1]], i32 1
; CHECK-NEXT: br label %[[BB1]]
;
bb:
br label %bb1

bb1:
%phi = phi i32 [ 0, %bb ], [ %mul9, %bb1 ]
%phi2 = phi i32 [ 0, %bb ], [ 0, %bb1 ]
%trunc = trunc i64 0 to i32
%mul = mul i32 0, %trunc
%mul3 = mul i32 %trunc, %phi
%mul4 = mul i32 %mul3, %mul
%mul5 = mul i32 %mul4, %mul
%trunc6 = trunc i64 0 to i32
%mul7 = mul i32 0, %trunc6
%mul8 = mul i32 %mul5, %mul7
%mul9 = mul i32 %mul8, %mul7
br label %bb1
}