Skip to content

[InstCombine] Pull extract through broadcast #143380

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 13 commits into from
Jul 4, 2025
Merged

[InstCombine] Pull extract through broadcast #143380

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
34c1222
Initial changes
agorenstein-nvidia Jun 6, 2025
31cb512
Merge branch 'main' into instcombine-extract-through-broadcast
agorenstein-nvidia Jun 9, 2025
e315c89
PR feedback 1
agorenstein-nvidia Jun 9, 2025
274cec9
clang-format
agorenstein-nvidia Jun 9, 2025
4fda162
Fixing edit damage in test file
agorenstein-nvidia Jun 9, 2025
5d9e4d9
Removed unneeded check
agorenstein-nvidia Jun 9, 2025
a1aec2e
Improve undef/poison usage in test (per PR automation failure), and u…
agorenstein-nvidia Jun 9, 2025
70d3c01
Fixed varname convention per PR feedback; on reflection from PR feedb…
agorenstein-nvidia Jun 9, 2025
52f5c4b
Restore old tests, reframe optimization as extension of existing one.
agorenstein-nvidia Jun 10, 2025
e21bde6
fix formatting errors, sorry
agorenstein-nvidia Jun 16, 2025
09d5e4c
Improvement via optional from RKSimon's suggestion
agorenstein-nvidia Jun 24, 2025
ce809f6
Improved tests with semantic names per RKSimon's PR feedback
agorenstein-nvidia Jun 24, 2025
21b8638
Improved comments to explicitly-named tests per PR suggestion
agorenstein-nvidia Jun 26, 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
7 changes: 7 additions & 0 deletions llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -542,6 +542,13 @@ Instruction *InstCombinerImpl::visitExtractElementInst(ExtractElementInst &EI) {
}
}
} else if (auto *SVI = dyn_cast<ShuffleVectorInst>(I)) {
// extractelt (shufflevector %v1, %v2, splat-mask) idx ->
// extractelt %v1, splat-mask[0]
auto mask = SVI->getShuffleMask();
if (mask[0] != PoisonMaskElem && all_equal(mask))
return ExtractElementInst::Create(SVI->getOperand(0),
Builder.getInt64(mask[0]));

// If this is extracting an element from a shufflevector, figure out where
// it came from and extract from the appropriate input element instead.
// Restrict the following transformation to fixed-length vector.
Expand Down
54 changes: 54 additions & 0 deletions llvm/test/Transforms/InstCombine/vec_extract_through_broadcast.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,54 @@
; RUN: opt -passes=instcombine -S < %s | FileCheck %s

define float @extract_from_zero_init_shuffle(<2 x float> %1, i64 %idx) {
; CHECK-LABEL: @extract_from_zero_init_shuffle(
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x float> [[W:%.*]], i64 0
; CHECK-NEXT: ret float [[TMP1]]
;
%3 = shufflevector <2 x float> %1, <2 x float> poison, <4 x i32> zeroinitializer
Copy link
Collaborator

@RKSimon RKSimon Jun 24, 2025

Choose a reason for hiding this comment

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

(style) avoid using numbers for variable names (%vec, %shuffle and %extract would be better)

Copy link
Contributor Author

@agorenstein-nvidia agorenstein-nvidia Jun 24, 2025

Choose a reason for hiding this comment

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

Done in latest push, please take a look. Thanks.

%4 = extractelement <4 x float> %3, i64 %idx
ret float %4
}


define float @extract_from_general_splat(<2 x float> %1, i64 %idx) {
; CHECK-LABEL: @extract_from_general_splat(
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x float> [[W:%.*]], i64 1
; CHECK-NEXT: ret float [[TMP1]]
;
%3 = shufflevector <2 x float> %1, <2 x float> poison, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
%4 = extractelement <4 x float> %3, i64 %idx
ret float %4
}

define float @extract_from_general_scalable_splat(<vscale x 2 x float> %1, i64 %idx) {
; CHECK-LABEL: @extract_from_general_scalable_splat(
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <vscale x 2 x float> [[W:%.*]], i64 0
; CHECK-NEXT: ret float [[TMP1]]
;
%3 = shufflevector <vscale x 2 x float> %1, <vscale x 2 x float> poison, <vscale x 4 x i32> zeroinitializer
%4 = extractelement <vscale x 4 x float> %3, i64 %idx
ret float %4
}

define float @no_extract_from_general_no_splat_0(<2 x float> %1, i64 %idx) {
; CHECK-LABEL: @no_extract_from_general_no_splat_0(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x float> [[W:%.*]], <2 x float> poison, <4 x i32> <i32 poison, i32 1, i32 1, i32 1>
; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i64 %idx
; CHECK-NEXT: ret float [[TMP2]]
;
%3 = shufflevector <2 x float> %1, <2 x float> poison, <4 x i32> <i32 poison, i32 1, i32 1, i32 1>
%4 = extractelement <4 x float> %3, i64 %idx
ret float %4
}

define float @no_extract_from_general_no_splat_1(<2 x float> %1, i64 %idx) {
; CHECK-LABEL: @no_extract_from_general_no_splat_1(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x float> [[W:%.*]], <2 x float> poison, <4 x i32> <i32 1, i32 poison, i32 1, i32 1>
; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[TMP1]], i64 %idx
; CHECK-NEXT: ret float [[TMP2]]
;
%3 = shufflevector <2 x float> %1, <2 x float> poison, <4 x i32> <i32 1, i32 poison, i32 1, i32 1>
%4 = extractelement <4 x float> %3, i64 %idx
ret float %4
}