Skip to content

[LLVM][CodeGen][AArch64] Don't scalarise v8{f16,bf16} vsetcc operations. #135398

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
May 1, 2025
Merged

[LLVM][CodeGen][AArch64] Don't scalarise v8{f16,bf16} vsetcc operations. #135398

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
59a2572
[LLVM][CodeGen][AArch64] Don't scalarise v8{f16,bf16} vsetcc operations.
paulwalker-arm Apr 11, 2025
079a9a1
Add comment to assert.
paulwalker-arm Apr 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
37 changes: 12 additions & 25 deletions llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -843,18 +843,21 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
setOperationPromotedToType(ISD::FRINT, V4Narrow, MVT::v4f32);
setOperationPromotedToType(ISD::FNEARBYINT, V4Narrow, MVT::v4f32);
Copy link
Collaborator

Choose a reason for hiding this comment

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

I would expect this to just mark the node as promote. Do you know why this needs to provide the necessary vector type to promote to, and that isn't automatically handled by getTypeToPromoteTo?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

getTypeToPromoteTo does not work in the natural way when operating on vector types. Specially, there is no guarantee the input and output element counts will match and indeed on AArch64 getTypeToPromoteTo(v4f16) returns v2f32.

Copy link
Collaborator

Choose a reason for hiding this comment

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

Sorry I missed this patch in other reviews and reviewers - my point was that perhaps it is better to get getTypeToPromoteTo to work for vectors too. It doesn't seem very useful to have the function never work and every target need to override half and bfloat vectors types. Something to think about at least, the patch looks good.

Copy link
Collaborator Author

@paulwalker-arm paulwalker-arm May 1, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

Thanks, I'll add it to my TODO list to revisit. I have investigated this in the past and found uses that relied on the current behaviour. I recall things like using promotion to canonicalise all vector loads to integer vector loads. Perhaps I misunderstood the code back then or the compiler has simply moved on. Hopefully the latter because like you say the current default doesn't feel much like promotion.

setOperationPromotedToType(ISD::FCANONICALIZE, V4Narrow, MVT::v4f32);
setOperationPromotedToType(ISD::SETCC, V4Narrow, MVT::v4f32);

Copy link
Collaborator Author

@paulwalker-arm paulwalker-arm Apr 11, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

This is a tab removal.

setOperationAction(ISD::FABS, V4Narrow, Legal);
setOperationAction(ISD::FNEG, V4Narrow, Legal);
setOperationAction(ISD::FNEG, V4Narrow, Legal);
setOperationAction(ISD::FMA, V4Narrow, Expand);
setOperationAction(ISD::SETCC, V4Narrow, Custom);
setOperationAction(ISD::BR_CC, V4Narrow, Expand);
setOperationAction(ISD::SELECT, V4Narrow, Expand);
setOperationAction(ISD::SELECT_CC, V4Narrow, Expand);
setOperationAction(ISD::FCOPYSIGN, V4Narrow, Custom);
setOperationAction(ISD::FSQRT, V4Narrow, Expand);

auto V8Narrow = MVT::getVectorVT(ScalarVT, 8);
setOperationPromotedToType(ISD::FCANONICALIZE, V8Narrow, MVT::v8f32);
setOperationPromotedToType(ISD::SETCC, V8Narrow, MVT::v8f32);

setOperationAction(ISD::FABS, V8Narrow, Legal);
setOperationAction(ISD::FADD, V8Narrow, Legal);
setOperationAction(ISD::FCEIL, V8Narrow, Legal);
Expand All @@ -864,19 +867,17 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
setOperationAction(ISD::FMA, V8Narrow, Expand);
setOperationAction(ISD::FMUL, V8Narrow, Legal);
setOperationAction(ISD::FNEARBYINT, V8Narrow, Legal);
setOperationAction(ISD::FNEG, V8Narrow, Legal);
setOperationAction(ISD::FNEG, V8Narrow, Legal);
setOperationAction(ISD::FROUND, V8Narrow, Legal);
setOperationAction(ISD::FROUNDEVEN, V8Narrow, Legal);
setOperationAction(ISD::FRINT, V8Narrow, Legal);
setOperationAction(ISD::FSQRT, V8Narrow, Expand);
setOperationAction(ISD::FSUB, V8Narrow, Legal);
setOperationAction(ISD::FTRUNC, V8Narrow, Legal);
setOperationAction(ISD::SETCC, V8Narrow, Expand);
setOperationAction(ISD::BR_CC, V8Narrow, Expand);
setOperationAction(ISD::SELECT, V8Narrow, Expand);
setOperationAction(ISD::SELECT_CC, V8Narrow, Expand);
setOperationAction(ISD::FP_EXTEND, V8Narrow, Expand);
setOperationPromotedToType(ISD::FCANONICALIZE, V8Narrow, MVT::v8f32);
};

if (!Subtarget->hasFullFP16()) {
Expand Down Expand Up @@ -15904,6 +15905,12 @@ SDValue AArch64TargetLowering::LowerVSETCC(SDValue Op,
if (LHS.getValueType().getVectorElementType().isInteger())
return Op;

assert(((!Subtarget->hasFullFP16() &&
LHS.getValueType().getVectorElementType() != MVT::f16) ||
LHS.getValueType().getVectorElementType() != MVT::bf16 ||
LHS.getValueType().getVectorElementType() != MVT::f128) &&
"Unexpected type!");

// Lower isnan(x) | isnan(never-nan) to x != x.
// Lower !isnan(x) & !isnan(never-nan) to x == x.
if (CC == ISD::SETUO || CC == ISD::SETO) {
Expand All @@ -15922,26 +15929,6 @@ SDValue AArch64TargetLowering::LowerVSETCC(SDValue Op,
}
}

const bool FullFP16 = DAG.getSubtarget<AArch64Subtarget>().hasFullFP16();

// Make v4f16 (only) fcmp operations utilise vector instructions
// v8f16 support will be a litle more complicated
if ((!FullFP16 && LHS.getValueType().getVectorElementType() == MVT::f16) ||
LHS.getValueType().getVectorElementType() == MVT::bf16) {
if (LHS.getValueType().getVectorNumElements() == 4) {
LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::v4f32, LHS);
RHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::v4f32, RHS);
SDValue NewSetcc = DAG.getSetCC(dl, MVT::v4i16, LHS, RHS, CC);
DAG.ReplaceAllUsesWith(Op, NewSetcc);
CmpVT = MVT::v4i32;
} else
return SDValue();
}

assert((!FullFP16 && LHS.getValueType().getVectorElementType() != MVT::f16) ||
LHS.getValueType().getVectorElementType() != MVT::bf16 ||
LHS.getValueType().getVectorElementType() != MVT::f128);

// Unfortunately, the mapping of LLVM FP CC's onto AArch64 CC's isn't totally
// clean. Some of them require two branches to implement.
AArch64CC::CondCode CC1, CC2;
Expand Down
5 changes: 0 additions & 5 deletions llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4241,11 +4241,6 @@ InstructionCost AArch64TTIImpl::getCmpSelInstrCost(
ValScalarTy->isBFloatTy()) {
auto *ValVTy = cast<FixedVectorType>(ValTy);

// FIXME: We currently scalarise these.
if (ValVTy->getNumElements() > 4)
return BaseT::getCmpSelInstrCost(Opcode, ValTy, CondTy, VecPred,
CostKind, Op1Info, Op2Info, I);

// Without dedicated instructions we promote [b]f16 compares to f32.
auto *PromotedTy =
VectorType::get(Type::getFloatTy(ValTy->getContext()), ValVTy);
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/Analysis/CostModel/AArch64/cmp.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,10 +17,10 @@ define void @cmps() {
; CHECK-NEXT: Cost Model: Found costs of 1 for: %cf32 = fcmp ogt float undef, undef
; CHECK-NEXT: Cost Model: Found costs of 1 for: %cf64 = fcmp ogt double undef, undef
; CHECK-NEXT: Cost Model: Found costs of 1 for: %cbf64 = fcmp ogt bfloat undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cfv816 = fcmp olt <8 x half> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cfv816 = fcmp olt <8 x half> undef, undef
; CHECK-NEXT: Cost Model: Found costs of 1 for: %cfv432 = fcmp oge <4 x float> undef, undef
; CHECK-NEXT: Cost Model: Found costs of 1 for: %cfv264 = fcmp oge <2 x double> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cbfv816 = fcmp olt <8 x bfloat> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cbfv816 = fcmp olt <8 x bfloat> undef, undef
; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
%c8 = icmp slt i8 undef, undef
Expand Down
32 changes: 16 additions & 16 deletions llvm/test/Analysis/CostModel/AArch64/vector-select.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,7 +119,7 @@ define <4 x half> @v4f16_select_ogt(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_ogt(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_ogt'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ogt <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ogt <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -179,7 +179,7 @@ define <4 x bfloat> @v4bf16_select_ogt(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_ogt(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_ogt'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ogt <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ogt <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -206,7 +206,7 @@ define <4 x half> @v4f16_select_oge(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_oge(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_oge'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oge <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oge <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -266,7 +266,7 @@ define <4 x bfloat> @v4bf16_select_oge(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_oge(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_oge'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oge <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oge <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -293,7 +293,7 @@ define <4 x half> @v4f16_select_olt(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_olt(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_olt'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp olt <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp olt <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -353,7 +353,7 @@ define <4 x bfloat> @v4bf16_select_olt(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_olt(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_olt'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp olt <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp olt <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -380,7 +380,7 @@ define <4 x half> @v4f16_select_ole(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_ole(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_ole'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ole <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ole <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -440,7 +440,7 @@ define <4 x bfloat> @v4bf16_select_ole(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_ole(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_ole'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ole <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ole <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -467,7 +467,7 @@ define <4 x half> @v4f16_select_oeq(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_oeq(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_oeq'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oeq <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oeq <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -527,7 +527,7 @@ define <4 x bfloat> @v4bf16_select_oeq(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_oeq(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_oeq'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oeq <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp oeq <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -554,7 +554,7 @@ define <4 x half> @v4f16_select_one(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_one(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_one'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp one <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp one <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -614,7 +614,7 @@ define <4 x bfloat> @v4bf16_select_one(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_one(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_one'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp one <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp one <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -641,7 +641,7 @@ define <4 x half> @v4f16_select_une(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_une(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_une'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp une <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp une <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -701,7 +701,7 @@ define <4 x bfloat> @v4bf16_select_une(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_une(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_une'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp une <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp une <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand All @@ -728,7 +728,7 @@ define <4 x half> @v4f16_select_ord(<4 x half> %a, <4 x half> %b, <4 x half> %c)

define <8 x half> @v8f16_select_ord(<8 x half> %a, <8 x half> %b, <8 x half> %c) {
; COST-NOFP16-LABEL: 'v8f16_select_ord'
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ord <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ord <8 x half> %a, %b
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x half> %a, <8 x half> %c
; COST-NOFP16-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x half> %s.1
;
Expand Down Expand Up @@ -788,7 +788,7 @@ define <4 x bfloat> @v4bf16_select_ord(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bf

define <8 x bfloat> @v8bf16_select_ord(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
; COST-LABEL: 'v8bf16_select_ord'
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ord <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:7 CodeSize:1 Lat:1 SizeLat:1 for: %cmp.1 = fcmp ord <8 x bfloat> %a, %b
; COST-NEXT: Cost Model: Found costs of RThru:22 CodeSize:1 Lat:1 SizeLat:1 for: %s.1 = select <8 x i1> %cmp.1, <8 x bfloat> %a, <8 x bfloat> %c
; COST-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x bfloat> %s.1
;
Expand Down
Loading
Loading