Skip to content

[InstCombine] Pull unary shuffles through fneg/fabs #144933

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
Jun 30, 2025
+53 −66
Merged

[InstCombine] Pull unary shuffles through fneg/fabs #144933

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
4bdc3d6
[InstCombine] Pull unary shuffles through fneg/fabs
lukel97 Jun 19, 2025
509f3aa
Merge branch 'main' into instcombine/fneg-fabs-of-shuffle
lukel97 Jun 30, 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
5 changes: 5 additions & 0 deletions llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3044,6 +3044,11 @@ Instruction *InstCombinerImpl::visitFNeg(UnaryOperator &I) {
return replaceInstUsesWith(I, NewCopySign);
}

// fneg (shuffle x, Mask) --> shuffle (fneg x), Mask
ArrayRef<int> Mask;
if (match(OneUse, m_Shuffle(m_Value(X), m_Poison(), m_Mask(Mask))))
return new ShuffleVectorInst(Builder.CreateFNegFMF(X, &I), Mask);

return nullptr;
}

Expand Down
5 changes: 0 additions & 5 deletions llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1404,11 +1404,6 @@ InstCombinerImpl::foldShuffledIntrinsicOperands(IntrinsicInst *II) {
!II->getCalledFunction()->isSpeculatable())
return nullptr;

// fabs is canonicalized to fabs (shuffle ...) in foldShuffleOfUnaryOps, so
// avoid undoing it.
if (match(II, m_FAbs(m_Value())))
return nullptr;

Value *X;
Constant *C;
ArrayRef<int> Mask;
Expand Down
14 changes: 0 additions & 14 deletions llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2533,20 +2533,6 @@ static Instruction *foldShuffleOfUnaryOps(ShuffleVectorInst &Shuf,

bool IsFNeg = S0->getOpcode() == Instruction::FNeg;

// Match 1-input (unary) shuffle.
// shuffle (fneg/fabs X), Mask --> fneg/fabs (shuffle X, Mask)
if (S0->hasOneUse() && match(Shuf.getOperand(1), m_Poison())) {
Value *NewShuf = Builder.CreateShuffleVector(X, Shuf.getShuffleMask());
if (IsFNeg)
return UnaryOperator::CreateFNegFMF(NewShuf, S0);

Function *FAbs = Intrinsic::getOrInsertDeclaration(
Shuf.getModule(), Intrinsic::fabs, Shuf.getType());
CallInst *NewF = CallInst::Create(FAbs, {NewShuf});
NewF->setFastMathFlags(S0->getFastMathFlags());
return NewF;
}

// Match 2-input (binary) shuffle.
auto *S1 = dyn_cast<Instruction>(Shuf.getOperand(1));
Value *Y;
Expand Down
7 changes: 4 additions & 3 deletions llvm/test/Transforms/InstCombine/matrix-multiplication-negation.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,9 +205,10 @@ define <4 x double> @matrix_multiply_two_operands_negated_with_same_size(<2 x do

define <2 x double> @matrix_multiply_two_operands_with_multiple_uses(<6 x double> %a, <3 x double> %b) {
; CHECK-LABEL: @matrix_multiply_two_operands_with_multiple_uses(
; CHECK-NEXT: [[RES:%.*]] = tail call <2 x double> @llvm.matrix.multiply.v2f64.v6f64.v3f64(<6 x double> [[A:%.*]], <3 x double> [[B:%.*]], i32 2, i32 3, i32 1)
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <6 x double> [[A]], <6 x double> poison, <2 x i32> <i32 0, i32 1>
; CHECK-NEXT: [[RES_3:%.*]] = fsub <2 x double> [[RES]], [[TMP1]]
; CHECK-NEXT: [[A_NEG:%.*]] = fneg <6 x double> [[A:%.*]]
; CHECK-NEXT: [[RES:%.*]] = tail call <2 x double> @llvm.matrix.multiply.v2f64.v6f64.v3f64(<6 x double> [[A]], <3 x double> [[B:%.*]], i32 2, i32 3, i32 1)
; CHECK-NEXT: [[RES_2:%.*]] = shufflevector <6 x double> [[A_NEG]], <6 x double> poison, <2 x i32> <i32 0, i32 1>
; CHECK-NEXT: [[RES_3:%.*]] = fadd <2 x double> [[RES_2]], [[RES]]
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 don't really have a good answer for this diff other than that it still seems to test what its supposed to test, i.e. that the operands to the intrinsic get swapped out with their non-negative counterparts.

; CHECK-NEXT: ret <2 x double> [[RES_3]]
;
%a.neg = fneg <6 x double> %a
Expand Down
6 changes: 3 additions & 3 deletions llvm/test/Transforms/InstCombine/vec_shuffle-inseltpoison.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1278,9 +1278,9 @@ define <2 x float> @fsub_splat_constant1(<2 x float> %x) {

define <2 x float> @fneg(<2 x float> %x) {
; CHECK-LABEL: @fneg(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x float> [[X:%.*]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[R:%.*]] = fneg <2 x float> [[TMP1]]
; CHECK-NEXT: ret <2 x float> [[R]]
; CHECK-NEXT: [[R:%.*]] = fneg <2 x float> [[TMP1:%.*]]
; CHECK-NEXT: [[R1:%.*]] = shufflevector <2 x float> [[R]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: ret <2 x float> [[R1]]
;
%splat = shufflevector <2 x float> %x, <2 x float> poison, <2 x i32> zeroinitializer
%r = fsub <2 x float> <float -0.0, float -0.0>, %splat
Expand Down
42 changes: 21 additions & 21 deletions llvm/test/Transforms/InstCombine/vec_shuffle.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1382,9 +1382,9 @@ define <2 x float> @fsub_splat_constant1(<2 x float> %x) {

define <2 x float> @fneg(<2 x float> %x) {
; CHECK-LABEL: @fneg(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x float> [[X:%.*]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[R:%.*]] = fneg <2 x float> [[TMP1]]
; CHECK-NEXT: ret <2 x float> [[R]]
; CHECK-NEXT: [[R:%.*]] = fneg <2 x float> [[TMP1:%.*]]
; CHECK-NEXT: [[R1:%.*]] = shufflevector <2 x float> [[R]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: ret <2 x float> [[R1]]
;
%splat = shufflevector <2 x float> %x, <2 x float> undef, <2 x i32> zeroinitializer
%r = fsub <2 x float> <float -0.0, float -0.0>, %splat
Expand Down Expand Up @@ -1906,9 +1906,9 @@ define <4 x i32> @PR46872(<4 x i32> %x) {

define <2 x float> @fabs_unary_shuf(<2 x float> %x) {
; CHECK-LABEL: @fabs_unary_shuf(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x float> [[X:%.*]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT: [[R:%.*]] = call nnan nsz <2 x float> @llvm.fabs.v2f32(<2 x float> [[TMP1]])
; CHECK-NEXT: ret <2 x float> [[R]]
; CHECK-NEXT: [[R:%.*]] = call nnan nsz <2 x float> @llvm.fabs.v2f32(<2 x float> [[TMP1:%.*]])
; CHECK-NEXT: [[R1:%.*]] = shufflevector <2 x float> [[R]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT: ret <2 x float> [[R1]]
;
%nx = call nsz nnan <2 x float> @llvm.fabs.v2f32(<2 x float> %x)
%r = shufflevector <2 x float> %nx, <2 x float> poison, <2 x i32> <i32 1, i32 0>
Expand All @@ -1917,9 +1917,9 @@ define <2 x float> @fabs_unary_shuf(<2 x float> %x) {

define <4 x half> @fabs_unary_shuf_widen(<2 x half> %x) {
; CHECK-LABEL: @fabs_unary_shuf_widen(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x half> [[X:%.*]], <2 x half> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 poison>
; CHECK-NEXT: [[R:%.*]] = call ninf <4 x half> @llvm.fabs.v4f16(<4 x half> [[TMP1]])
; CHECK-NEXT: ret <4 x half> [[R]]
; CHECK-NEXT: [[X:%.*]] = call ninf <2 x half> @llvm.fabs.v2f16(<2 x half> [[X1:%.*]])
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x half> [[X]], <2 x half> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 poison>
; CHECK-NEXT: ret <4 x half> [[TMP1]]
;
%nx = call ninf <2 x half> @llvm.fabs.v2f16(<2 x half> %x)
%r = shufflevector <2 x half> %nx, <2 x half> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 poison>
Expand All @@ -1928,9 +1928,9 @@ define <4 x half> @fabs_unary_shuf_widen(<2 x half> %x) {

define <2 x double> @fabs_unary_shuf_narrow(<4 x double> %x) {
; CHECK-LABEL: @fabs_unary_shuf_narrow(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[X:%.*]], <4 x double> poison, <2 x i32> <i32 3, i32 0>
; CHECK-NEXT: [[R:%.*]] = call nsz <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP1]])
; CHECK-NEXT: ret <2 x double> [[R]]
; CHECK-NEXT: [[X:%.*]] = call nsz <4 x double> @llvm.fabs.v4f64(<4 x double> [[X1:%.*]])
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[X]], <4 x double> poison, <2 x i32> <i32 3, i32 0>
; CHECK-NEXT: ret <2 x double> [[TMP1]]
;
%nx = call nsz <4 x double> @llvm.fabs.v4f64(<4 x double> %x)
%r = shufflevector <4 x double> %nx, <4 x double> poison, <2 x i32> <i32 3, i32 0>
Expand Down Expand Up @@ -2021,9 +2021,9 @@ define <2 x float> @fabs_shuf_use3(<2 x float> %x, <2 x float> %y) {

define <2 x float> @fneg_unary_shuf(<2 x float> %x) {
; CHECK-LABEL: @fneg_unary_shuf(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x float> [[X:%.*]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT: [[R:%.*]] = fneg nnan nsz <2 x float> [[TMP1]]
; CHECK-NEXT: ret <2 x float> [[R]]
; CHECK-NEXT: [[R:%.*]] = fneg nnan nsz <2 x float> [[TMP1:%.*]]
; CHECK-NEXT: [[R1:%.*]] = shufflevector <2 x float> [[R]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT: ret <2 x float> [[R1]]
;
%nx = fneg nsz nnan <2 x float> %x
%r = shufflevector <2 x float> %nx, <2 x float> poison, <2 x i32> <i32 1, i32 0>
Expand All @@ -2032,9 +2032,9 @@ define <2 x float> @fneg_unary_shuf(<2 x float> %x) {

define <4 x half> @fneg_unary_shuf_widen(<2 x half> %x) {
; CHECK-LABEL: @fneg_unary_shuf_widen(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x half> [[X:%.*]], <2 x half> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 poison>
; CHECK-NEXT: [[R:%.*]] = fneg ninf <4 x half> [[TMP1]]
; CHECK-NEXT: ret <4 x half> [[R]]
; CHECK-NEXT: [[X:%.*]] = fneg ninf <2 x half> [[X1:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x half> [[X]], <2 x half> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 poison>
; CHECK-NEXT: ret <4 x half> [[TMP1]]
;
%nx = fneg ninf <2 x half> %x
%r = shufflevector <2 x half> %nx, <2 x half> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 poison>
Expand All @@ -2043,9 +2043,9 @@ define <4 x half> @fneg_unary_shuf_widen(<2 x half> %x) {

define <2 x double> @fneg_unary_shuf_narrow(<4 x double> %x) {
; CHECK-LABEL: @fneg_unary_shuf_narrow(
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[X:%.*]], <4 x double> poison, <2 x i32> <i32 3, i32 0>
; CHECK-NEXT: [[R:%.*]] = fneg nsz <2 x double> [[TMP1]]
; CHECK-NEXT: ret <2 x double> [[R]]
; CHECK-NEXT: [[X:%.*]] = fneg nsz <4 x double> [[X1:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x double> [[X]], <4 x double> poison, <2 x i32> <i32 3, i32 0>
; CHECK-NEXT: ret <2 x double> [[TMP1]]
;
%nx = fneg nsz <4 x double> %x
%r = shufflevector <4 x double> %nx, <4 x double> poison, <2 x i32> <i32 3, i32 0>
Expand Down
20 changes: 10 additions & 10 deletions llvm/test/Transforms/SLPVectorizer/X86/alternate-cast-inseltpoison.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,19 +132,19 @@ define <8 x i32> @fptosi_fptoui(<8 x float> %a) {

define <8 x float> @fneg_fabs(<8 x float> %a) {
; SSE2-LABEL: @fneg_fabs(
; SSE2-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SSE2-NEXT: [[TMP2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: [[TMP3:%.*]] = fneg <4 x float> [[TMP1]]
; SSE2-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP2]])
; SSE2-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: [[A:%.*]] = fneg <8 x float> [[A1:%.*]]
; SSE2-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SSE2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[A1]])
; SSE2-NEXT: [[TMP4:%.*]] = shufflevector <8 x float> [[TMP3]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
Copy link
Contributor Author

Choose a reason for hiding this comment

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

By pulling the shufflevectors through VectorCombine can now fold them away so we're left with just one shuffle

Copy link
Collaborator

Choose a reason for hiding this comment

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

There are a lot of slp-vectorizer tests that currently run as -passes=slp-vectorizer,instcombine - but as vector-combine typically runs between these passes, maybe we should change these to -passes=slp-vectorizer,vector-combine,instcombine @alexey-bataev WDYT? An alternative is we move them to phase-ordering.

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 just double checked -O3 -print-pipeline-passes and that is indeed the order. I wonder though should we be running instcombine before vector-combine too to see if it helps more shuffles get folded?

Copy link
Member

Choose a reason for hiding this comment

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

There are a lot of slp-vectorizer tests that currently run as -passes=slp-vectorizer,instcombine - but as vector-combine typically runs between these passes, maybe we should change these to -passes=slp-vectorizer,vector-combine,instcombine @alexey-bataev WDYT? An alternative is we move them to phase-ordering.

I would actually remove instcombine from slp tests, where possible, it makes it harder to estimate transformations

; SSE2-NEXT: ret <8 x float> [[DOTUNCASTED]]
;
; SLM-LABEL: @fneg_fabs(
; SLM-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SLM-NEXT: [[TMP2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: [[TMP3:%.*]] = fneg <4 x float> [[TMP1]]
; SLM-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP2]])
; SLM-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: [[A:%.*]] = fneg <8 x float> [[A1:%.*]]
; SLM-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SLM-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[A1]])
; SLM-NEXT: [[TMP4:%.*]] = shufflevector <8 x float> [[TMP3]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: ret <8 x float> [[DOTUNCASTED]]
;
; AVX-LABEL: @fneg_fabs(
Expand Down
20 changes: 10 additions & 10 deletions llvm/test/Transforms/SLPVectorizer/X86/alternate-cast.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,19 +132,19 @@ define <8 x i32> @fptosi_fptoui(<8 x float> %a) {

define <8 x float> @fneg_fabs(<8 x float> %a) {
; SSE2-LABEL: @fneg_fabs(
; SSE2-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SSE2-NEXT: [[TMP2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: [[TMP3:%.*]] = fneg <4 x float> [[TMP1]]
; SSE2-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP2]])
; SSE2-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: [[A:%.*]] = fneg <8 x float> [[A1:%.*]]
; SSE2-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SSE2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[A1]])
; SSE2-NEXT: [[TMP4:%.*]] = shufflevector <8 x float> [[TMP3]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SSE2-NEXT: ret <8 x float> [[DOTUNCASTED]]
;
; SLM-LABEL: @fneg_fabs(
; SLM-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SLM-NEXT: [[TMP2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: [[TMP3:%.*]] = fneg <4 x float> [[TMP1]]
; SLM-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP2]])
; SLM-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: [[A:%.*]] = fneg <8 x float> [[A1:%.*]]
; SLM-NEXT: [[TMP1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; SLM-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[A1]])
; SLM-NEXT: [[TMP4:%.*]] = shufflevector <8 x float> [[TMP3]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: [[DOTUNCASTED:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; SLM-NEXT: ret <8 x float> [[DOTUNCASTED]]
;
; AVX-LABEL: @fneg_fabs(
Expand Down
Loading