[AMDGPU] Fold llvm.amdgcn.cvt.pkrtz when either operand is fpext #108237
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This also generalizes the Undef handling and adds Poison handling.
|
@llvm/pr-subscribers-backend-amdgpu Author: Jay Foad (jayfoad) ChangesThis also generalizes the Undef handling and adds Poison handling. Full diff: https://github.com/llvm/llvm-project/pull/108237.diff 2 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
index 4da3618357c420..389cc35b51f9da 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
@@ -643,27 +643,38 @@ GCNTTIImpl::instCombineIntrinsic(InstCombiner &IC, IntrinsicInst &II) const {
break;
}
case Intrinsic::amdgcn_cvt_pkrtz: {
- Value *Src0 = II.getArgOperand(0);
- Value *Src1 = II.getArgOperand(1);
- if (const ConstantFP *C0 = dyn_cast<ConstantFP>(Src0)) {
- if (const ConstantFP *C1 = dyn_cast<ConstantFP>(Src1)) {
- const fltSemantics &HalfSem =
- II.getType()->getScalarType()->getFltSemantics();
+ auto foldFPTruncToF16RTZ = [](Value *Arg) -> Value * {
+ Type *HalfTy = Type::getHalfTy(Arg->getContext());
+
+ if (isa<PoisonValue>(Arg))
+ return PoisonValue::get(HalfTy);
+ if (isa<UndefValue>(Arg))
+ return UndefValue::get(HalfTy);
+
+ ConstantFP *CFP = nullptr;
+ if (match(Arg, m_ConstantFP(CFP))) {
bool LosesInfo;
- APFloat Val0 = C0->getValueAPF();
- APFloat Val1 = C1->getValueAPF();
- Val0.convert(HalfSem, APFloat::rmTowardZero, &LosesInfo);
- Val1.convert(HalfSem, APFloat::rmTowardZero, &LosesInfo);
-
- Constant *Folded =
- ConstantVector::get({ConstantFP::get(II.getContext(), Val0),
- ConstantFP::get(II.getContext(), Val1)});
- return IC.replaceInstUsesWith(II, Folded);
+ APFloat Val(CFP->getValueAPF());
+ Val.convert(APFloat::IEEEhalf(), APFloat::rmTowardZero, &LosesInfo);
+ return ConstantFP::get(HalfTy, Val);
}
- }
- if (isa<UndefValue>(Src0) && isa<UndefValue>(Src1)) {
- return IC.replaceInstUsesWith(II, UndefValue::get(II.getType()));
+ Value *Src = nullptr;
+ if (match(Arg, m_OneUse(m_FPExt(m_Value(Src))))) {
+ if (Src->getType()->isHalfTy())
+ return Src;
+ }
+
+ return nullptr;
+ };
+
+ if (Value *Src0 = foldFPTruncToF16RTZ(II.getArgOperand(0))) {
+ if (Value *Src1 = foldFPTruncToF16RTZ(II.getArgOperand(1))) {
+ Value *V = PoisonValue::get(II.getType());
+ V = IC.Builder.CreateInsertElement(V, Src0, (uint64_t)0);
+ V = IC.Builder.CreateInsertElement(V, Src1, (uint64_t)1);
+ return IC.replaceInstUsesWith(II, V);
+ }
}
break;
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
index f3a3b8c1dc5d8a..176b3560b4064d 100644
--- a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
@@ -1161,6 +1161,48 @@ define <2 x half> @constant_rtz_pkrtz() {
ret <2 x half> %cvt
}
+define <2 x half> @fpext_const_cvt_pkrtz(half %x) {
+; CHECK-LABEL: @fpext_const_cvt_pkrtz(
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> <half poison, half 0xH4200>, half [[X:%.*]], i64 0
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %ext = fpext half %x to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %ext, float 3.0)
+ ret <2 x half> %cvt
+}
+
+define <2 x half> @const_fpext_cvt_pkrtz(half %y) {
+; CHECK-LABEL: @const_fpext_cvt_pkrtz(
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> <half 0xH4500, half poison>, half [[Y:%.*]], i64 1
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %ext = fpext half %y to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float 5.0, float %ext)
+ ret <2 x half> %cvt
+}
+
+define <2 x half> @fpext_fpext_cvt_pkrtz(half %x, half %y) {
+; CHECK-LABEL: @fpext_fpext_cvt_pkrtz(
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x half> poison, half [[X:%.*]], i64 0
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> [[TMP1]], half [[Y:%.*]], i64 1
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %extx = fpext half %x to float
+ %exty = fpext half %y to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %extx, float %exty)
+ ret <2 x half> %cvt
+}
+
+define <2 x half> @poison_fpext_cvt_pkrtz(half %y) {
+; CHECK-LABEL: @poison_fpext_cvt_pkrtz(
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> poison, half [[Y:%.*]], i64 1
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %ext = fpext half %y to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float poison, float %ext)
+ ret <2 x half> %cvt
+}
+
; --------------------------------------------------------------------
; llvm.amdgcn.cvt.pknorm.i16
; --------------------------------------------------------------------
|
|
@llvm/pr-subscribers-llvm-transforms Author: Jay Foad (jayfoad) ChangesThis also generalizes the Undef handling and adds Poison handling. Full diff: https://github.com/llvm/llvm-project/pull/108237.diff 2 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
index 4da3618357c420..389cc35b51f9da 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
@@ -643,27 +643,38 @@ GCNTTIImpl::instCombineIntrinsic(InstCombiner &IC, IntrinsicInst &II) const {
break;
}
case Intrinsic::amdgcn_cvt_pkrtz: {
- Value *Src0 = II.getArgOperand(0);
- Value *Src1 = II.getArgOperand(1);
- if (const ConstantFP *C0 = dyn_cast<ConstantFP>(Src0)) {
- if (const ConstantFP *C1 = dyn_cast<ConstantFP>(Src1)) {
- const fltSemantics &HalfSem =
- II.getType()->getScalarType()->getFltSemantics();
+ auto foldFPTruncToF16RTZ = [](Value *Arg) -> Value * {
+ Type *HalfTy = Type::getHalfTy(Arg->getContext());
+
+ if (isa<PoisonValue>(Arg))
+ return PoisonValue::get(HalfTy);
+ if (isa<UndefValue>(Arg))
+ return UndefValue::get(HalfTy);
+
+ ConstantFP *CFP = nullptr;
+ if (match(Arg, m_ConstantFP(CFP))) {
bool LosesInfo;
- APFloat Val0 = C0->getValueAPF();
- APFloat Val1 = C1->getValueAPF();
- Val0.convert(HalfSem, APFloat::rmTowardZero, &LosesInfo);
- Val1.convert(HalfSem, APFloat::rmTowardZero, &LosesInfo);
-
- Constant *Folded =
- ConstantVector::get({ConstantFP::get(II.getContext(), Val0),
- ConstantFP::get(II.getContext(), Val1)});
- return IC.replaceInstUsesWith(II, Folded);
+ APFloat Val(CFP->getValueAPF());
+ Val.convert(APFloat::IEEEhalf(), APFloat::rmTowardZero, &LosesInfo);
+ return ConstantFP::get(HalfTy, Val);
}
- }
- if (isa<UndefValue>(Src0) && isa<UndefValue>(Src1)) {
- return IC.replaceInstUsesWith(II, UndefValue::get(II.getType()));
+ Value *Src = nullptr;
+ if (match(Arg, m_OneUse(m_FPExt(m_Value(Src))))) {
+ if (Src->getType()->isHalfTy())
+ return Src;
+ }
+
+ return nullptr;
+ };
+
+ if (Value *Src0 = foldFPTruncToF16RTZ(II.getArgOperand(0))) {
+ if (Value *Src1 = foldFPTruncToF16RTZ(II.getArgOperand(1))) {
+ Value *V = PoisonValue::get(II.getType());
+ V = IC.Builder.CreateInsertElement(V, Src0, (uint64_t)0);
+ V = IC.Builder.CreateInsertElement(V, Src1, (uint64_t)1);
+ return IC.replaceInstUsesWith(II, V);
+ }
}
break;
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
index f3a3b8c1dc5d8a..176b3560b4064d 100644
--- a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
@@ -1161,6 +1161,48 @@ define <2 x half> @constant_rtz_pkrtz() {
ret <2 x half> %cvt
}
+define <2 x half> @fpext_const_cvt_pkrtz(half %x) {
+; CHECK-LABEL: @fpext_const_cvt_pkrtz(
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> <half poison, half 0xH4200>, half [[X:%.*]], i64 0
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %ext = fpext half %x to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %ext, float 3.0)
+ ret <2 x half> %cvt
+}
+
+define <2 x half> @const_fpext_cvt_pkrtz(half %y) {
+; CHECK-LABEL: @const_fpext_cvt_pkrtz(
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> <half 0xH4500, half poison>, half [[Y:%.*]], i64 1
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %ext = fpext half %y to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float 5.0, float %ext)
+ ret <2 x half> %cvt
+}
+
+define <2 x half> @fpext_fpext_cvt_pkrtz(half %x, half %y) {
+; CHECK-LABEL: @fpext_fpext_cvt_pkrtz(
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x half> poison, half [[X:%.*]], i64 0
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> [[TMP1]], half [[Y:%.*]], i64 1
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %extx = fpext half %x to float
+ %exty = fpext half %y to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %extx, float %exty)
+ ret <2 x half> %cvt
+}
+
+define <2 x half> @poison_fpext_cvt_pkrtz(half %y) {
+; CHECK-LABEL: @poison_fpext_cvt_pkrtz(
+; CHECK-NEXT: [[CVT:%.*]] = insertelement <2 x half> poison, half [[Y:%.*]], i64 1
+; CHECK-NEXT: ret <2 x half> [[CVT]]
+;
+ %ext = fpext half %y to float
+ %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float poison, float %ext)
+ ret <2 x half> %cvt
+}
+
; --------------------------------------------------------------------
; llvm.amdgcn.cvt.pknorm.i16
; --------------------------------------------------------------------
|
arsenm
reviewed
Sep 12, 2024
| if (isa<PoisonValue>(Arg)) | ||
| return PoisonValue::get(HalfTy); | ||
| if (isa<UndefValue>(Arg)) | ||
| return UndefValue::get(HalfTy); |
There was a problem hiding this comment.
usually we do undef -> qnan for FP folds (although I think this is overly conservative, and I assume is only to exclude snan bit patterns which isn't guaranteed to quiet anyway)
There was a problem hiding this comment.
I was following ConstantFoldCastInstruction which will fold (fptrunc undef) to undef.
| %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float poison, float %ext) | ||
| ret <2 x half> %cvt | ||
| } | ||
|
|
There was a problem hiding this comment.
Add some tests with bfloat sources. Also negative multi use test
There was a problem hiding this comment.
bfloat tests would be negative since the intrinsic only supports half.
As for multi use tests, I'm not sure if they should be negative. I know I used m_OneUse in the implementation, but now I think the optimization might be beneficial even with multiple uses.
| ; CHECK-NEXT: ret <2 x half> [[CVT]] | ||
| ; | ||
| %ext = fpext half %y to float | ||
| %cvt = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float poison, float %ext) |
arsenm
approved these changes
Sep 18, 2024
tmsri
pushed a commit
to tmsri/llvm-project
that referenced
this pull request
Sep 19, 2024
…m#108237) This also generalizes the Undef handling and adds Poison handling.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This also generalizes the Undef handling and adds Poison handling.