[LLVM][IR] Refactor ConstantFold:FoldBitCast to fully support vector ConstantInt/FP. #116787
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
llvmbot
added
llvm:instcombine
|
@llvm/pr-subscribers-llvm-ir Author: Paul Walker (paulwalker-arm) ChangesThis fixes the code quality issue reported in #111149. Full diff: https://github.com/llvm/llvm-project/pull/116787.diff 3 Files Affected:
diff --git a/llvm/lib/IR/ConstantFold.cpp b/llvm/lib/IR/ConstantFold.cpp
index 2dbc6785c08b9d..3a04f81315e6c6 100644
--- a/llvm/lib/IR/ConstantFold.cpp
+++ b/llvm/lib/IR/ConstantFold.cpp
@@ -71,50 +71,51 @@ static Constant *FoldBitCast(Constant *V, Type *DestTy) {
if (SrcTy == DestTy)
return V; // no-op cast
- // Handle casts from one vector constant to another. We know that the src
- // and dest type have the same size (otherwise its an illegal cast).
- if (VectorType *DestPTy = dyn_cast<VectorType>(DestTy)) {
- if (V->isAllOnesValue())
- return Constant::getAllOnesValue(DestTy);
+ if (V->isAllOnesValue())
+ return Constant::getAllOnesValue(DestTy);
+ // Handle ConstantInt -> ConstantFP
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
// Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts
// This allows for other simplifications (although some of them
// can only be handled by Analysis/ConstantFolding.cpp).
- if (!isa<VectorType>(SrcTy))
- if (isa<ConstantInt>(V) || isa<ConstantFP>(V))
- return ConstantExpr::getBitCast(ConstantVector::get(V), DestPTy);
- return nullptr;
- }
+ if (isa<VectorType>(DestTy) && !isa<VectorType>(SrcTy))
+ return ConstantExpr::getBitCast(ConstantVector::get(V), DestTy);
- // Handle integral constant input.
- if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
+ // Make sure dest type is compatible with the folded fp constant.
// See note below regarding the PPC_FP128 restriction.
- if (DestTy->isFloatingPointTy() && !DestTy->isPPC_FP128Ty())
- return ConstantFP::get(DestTy->getContext(),
- APFloat(DestTy->getFltSemantics(),
- CI->getValue()));
+ if (!DestTy->isFPOrFPVectorTy() || DestTy->isPPC_FP128Ty() ||
+ DestTy->getScalarSizeInBits() != SrcTy->getScalarSizeInBits())
+ return nullptr;
- // Otherwise, can't fold this (vector?)
- return nullptr;
+ return ConstantFP::get(
+ DestTy,
+ APFloat(DestTy->getScalarType()->getFltSemantics(), CI->getValue()));
}
- // Handle ConstantFP input: FP -> Integral.
+ // Handle ConstantFP -> ConstantInt
if (ConstantFP *FP = dyn_cast<ConstantFP>(V)) {
+ // Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts
+ // This allows for other simplifications (although some of them
+ // can only be handled by Analysis/ConstantFolding.cpp).
+ if (isa<VectorType>(DestTy) && !isa<VectorType>(SrcTy))
+ return ConstantExpr::getBitCast(ConstantVector::get(V), DestTy);
+
// PPC_FP128 is really the sum of two consecutive doubles, where the first
// double is always stored first in memory, regardless of the target
// endianness. The memory layout of i128, however, depends on the target
// endianness, and so we can't fold this without target endianness
// information. This should instead be handled by
// Analysis/ConstantFolding.cpp
- if (FP->getType()->isPPC_FP128Ty())
+ if (SrcTy->isPPC_FP128Ty())
return nullptr;
// Make sure dest type is compatible with the folded integer constant.
- if (!DestTy->isIntegerTy())
+ if (!DestTy->isIntOrIntVectorTy() ||
+ DestTy->getScalarSizeInBits() != SrcTy->getScalarSizeInBits())
return nullptr;
- return ConstantInt::get(FP->getContext(),
- FP->getValueAPF().bitcastToAPInt());
+ return ConstantInt::get(DestTy, FP->getValueAPF().bitcastToAPInt());
}
return nullptr;
diff --git a/llvm/test/Transforms/InstCombine/bitcast.ll b/llvm/test/Transforms/InstCombine/bitcast.ll
index ce5a4635a9b0a2..37d41de3e99911 100644
--- a/llvm/test/Transforms/InstCombine/bitcast.ll
+++ b/llvm/test/Transforms/InstCombine/bitcast.ll
@@ -1,5 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+; RUN: opt < %s -passes=instcombine -use-constant-fp-for-fixed-length-splat -use-constant-int-for-fixed-length-splat -S | FileCheck %s
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
target triple = "x86_64-apple-darwin10.0.0"
diff --git a/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll b/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll
index b475b8199541d5..a3ad5f01391422 100644
--- a/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll
+++ b/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -passes=instsimplify -S | FileCheck %s --check-prefixes=CHECK,CONSTVEC
-; RUN: opt < %s -passes=instsimplify -use-constant-fp-for-fixed-length-splat -use-constant-int-for-fixed-length-splat -S | FileCheck %s --check-prefixes=CHECK,CONSTSPLAT
+; RUN: opt < %s -passes=instsimplify -S | FileCheck %s
+; RUN: opt < %s -passes=instsimplify -use-constant-fp-for-fixed-length-splat -use-constant-int-for-fixed-length-splat -S | FileCheck %s
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-f64:32:64-v64:64:64-v128:128:128"
define <2 x i64> @test1() {
@@ -76,12 +76,8 @@ define <4 x i32> @test9(<1 x i64> %y) {
}
define <1 x i1> @test10() {
-; CONSTVEC-LABEL: @test10(
-; CONSTVEC-NEXT: [[RET:%.*]] = icmp eq <1 x i64> <i64 bitcast (<1 x double> splat (double 0xFFFFFFFFFFFFFFFF) to i64)>, zeroinitializer
-; CONSTVEC-NEXT: ret <1 x i1> [[RET]]
-;
-; CONSTSPLAT-LABEL: @test10(
-; CONSTSPLAT-NEXT: ret <1 x i1> zeroinitializer
+; CHECK-LABEL: @test10(
+; CHECK-NEXT: ret <1 x i1> zeroinitializer
;
%ret = icmp eq <1 x i64> <i64 bitcast (<1 x double> <double 0xFFFFFFFFFFFFFFFF> to i64)>, zeroinitializer
ret <1 x i1> %ret
@@ -282,11 +278,8 @@ define <16 x i8> @bitcast_constexpr_16i8_8i16_u256uuu256uu() {
}
define <1 x i32> @bitcast_constexpr_scalar_fp_to_vector_int() {
-; CONSTVEC-LABEL: @bitcast_constexpr_scalar_fp_to_vector_int(
-; CONSTVEC-NEXT: ret <1 x i32> splat (i32 1065353216)
-;
-; CONSTSPLAT-LABEL: @bitcast_constexpr_scalar_fp_to_vector_int(
-; CONSTSPLAT-NEXT: ret <1 x i32> bitcast (<1 x float> splat (float 1.000000e+00) to <1 x i32>)
+; CHECK-LABEL: @bitcast_constexpr_scalar_fp_to_vector_int(
+; CHECK-NEXT: ret <1 x i32> splat (i32 1065353216)
;
%res = bitcast float 1.0 to <1 x i32>
ret <1 x i32> %res
|
|
@llvm/pr-subscribers-llvm-transforms Author: Paul Walker (paulwalker-arm) ChangesThis fixes the code quality issue reported in #111149. Full diff: https://github.com/llvm/llvm-project/pull/116787.diff 3 Files Affected:
diff --git a/llvm/lib/IR/ConstantFold.cpp b/llvm/lib/IR/ConstantFold.cpp
index 2dbc6785c08b9d..3a04f81315e6c6 100644
--- a/llvm/lib/IR/ConstantFold.cpp
+++ b/llvm/lib/IR/ConstantFold.cpp
@@ -71,50 +71,51 @@ static Constant *FoldBitCast(Constant *V, Type *DestTy) {
if (SrcTy == DestTy)
return V; // no-op cast
- // Handle casts from one vector constant to another. We know that the src
- // and dest type have the same size (otherwise its an illegal cast).
- if (VectorType *DestPTy = dyn_cast<VectorType>(DestTy)) {
- if (V->isAllOnesValue())
- return Constant::getAllOnesValue(DestTy);
+ if (V->isAllOnesValue())
+ return Constant::getAllOnesValue(DestTy);
+ // Handle ConstantInt -> ConstantFP
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
// Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts
// This allows for other simplifications (although some of them
// can only be handled by Analysis/ConstantFolding.cpp).
- if (!isa<VectorType>(SrcTy))
- if (isa<ConstantInt>(V) || isa<ConstantFP>(V))
- return ConstantExpr::getBitCast(ConstantVector::get(V), DestPTy);
- return nullptr;
- }
+ if (isa<VectorType>(DestTy) && !isa<VectorType>(SrcTy))
+ return ConstantExpr::getBitCast(ConstantVector::get(V), DestTy);
- // Handle integral constant input.
- if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
+ // Make sure dest type is compatible with the folded fp constant.
// See note below regarding the PPC_FP128 restriction.
- if (DestTy->isFloatingPointTy() && !DestTy->isPPC_FP128Ty())
- return ConstantFP::get(DestTy->getContext(),
- APFloat(DestTy->getFltSemantics(),
- CI->getValue()));
+ if (!DestTy->isFPOrFPVectorTy() || DestTy->isPPC_FP128Ty() ||
+ DestTy->getScalarSizeInBits() != SrcTy->getScalarSizeInBits())
+ return nullptr;
- // Otherwise, can't fold this (vector?)
- return nullptr;
+ return ConstantFP::get(
+ DestTy,
+ APFloat(DestTy->getScalarType()->getFltSemantics(), CI->getValue()));
}
- // Handle ConstantFP input: FP -> Integral.
+ // Handle ConstantFP -> ConstantInt
if (ConstantFP *FP = dyn_cast<ConstantFP>(V)) {
+ // Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts
+ // This allows for other simplifications (although some of them
+ // can only be handled by Analysis/ConstantFolding.cpp).
+ if (isa<VectorType>(DestTy) && !isa<VectorType>(SrcTy))
+ return ConstantExpr::getBitCast(ConstantVector::get(V), DestTy);
+
// PPC_FP128 is really the sum of two consecutive doubles, where the first
// double is always stored first in memory, regardless of the target
// endianness. The memory layout of i128, however, depends on the target
// endianness, and so we can't fold this without target endianness
// information. This should instead be handled by
// Analysis/ConstantFolding.cpp
- if (FP->getType()->isPPC_FP128Ty())
+ if (SrcTy->isPPC_FP128Ty())
return nullptr;
// Make sure dest type is compatible with the folded integer constant.
- if (!DestTy->isIntegerTy())
+ if (!DestTy->isIntOrIntVectorTy() ||
+ DestTy->getScalarSizeInBits() != SrcTy->getScalarSizeInBits())
return nullptr;
- return ConstantInt::get(FP->getContext(),
- FP->getValueAPF().bitcastToAPInt());
+ return ConstantInt::get(DestTy, FP->getValueAPF().bitcastToAPInt());
}
return nullptr;
diff --git a/llvm/test/Transforms/InstCombine/bitcast.ll b/llvm/test/Transforms/InstCombine/bitcast.ll
index ce5a4635a9b0a2..37d41de3e99911 100644
--- a/llvm/test/Transforms/InstCombine/bitcast.ll
+++ b/llvm/test/Transforms/InstCombine/bitcast.ll
@@ -1,5 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+; RUN: opt < %s -passes=instcombine -use-constant-fp-for-fixed-length-splat -use-constant-int-for-fixed-length-splat -S | FileCheck %s
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
target triple = "x86_64-apple-darwin10.0.0"
diff --git a/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll b/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll
index b475b8199541d5..a3ad5f01391422 100644
--- a/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll
+++ b/llvm/test/Transforms/InstSimplify/bitcast-vector-fold.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -passes=instsimplify -S | FileCheck %s --check-prefixes=CHECK,CONSTVEC
-; RUN: opt < %s -passes=instsimplify -use-constant-fp-for-fixed-length-splat -use-constant-int-for-fixed-length-splat -S | FileCheck %s --check-prefixes=CHECK,CONSTSPLAT
+; RUN: opt < %s -passes=instsimplify -S | FileCheck %s
+; RUN: opt < %s -passes=instsimplify -use-constant-fp-for-fixed-length-splat -use-constant-int-for-fixed-length-splat -S | FileCheck %s
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-f64:32:64-v64:64:64-v128:128:128"
define <2 x i64> @test1() {
@@ -76,12 +76,8 @@ define <4 x i32> @test9(<1 x i64> %y) {
}
define <1 x i1> @test10() {
-; CONSTVEC-LABEL: @test10(
-; CONSTVEC-NEXT: [[RET:%.*]] = icmp eq <1 x i64> <i64 bitcast (<1 x double> splat (double 0xFFFFFFFFFFFFFFFF) to i64)>, zeroinitializer
-; CONSTVEC-NEXT: ret <1 x i1> [[RET]]
-;
-; CONSTSPLAT-LABEL: @test10(
-; CONSTSPLAT-NEXT: ret <1 x i1> zeroinitializer
+; CHECK-LABEL: @test10(
+; CHECK-NEXT: ret <1 x i1> zeroinitializer
;
%ret = icmp eq <1 x i64> <i64 bitcast (<1 x double> <double 0xFFFFFFFFFFFFFFFF> to i64)>, zeroinitializer
ret <1 x i1> %ret
@@ -282,11 +278,8 @@ define <16 x i8> @bitcast_constexpr_16i8_8i16_u256uuu256uu() {
}
define <1 x i32> @bitcast_constexpr_scalar_fp_to_vector_int() {
-; CONSTVEC-LABEL: @bitcast_constexpr_scalar_fp_to_vector_int(
-; CONSTVEC-NEXT: ret <1 x i32> splat (i32 1065353216)
-;
-; CONSTSPLAT-LABEL: @bitcast_constexpr_scalar_fp_to_vector_int(
-; CONSTSPLAT-NEXT: ret <1 x i32> bitcast (<1 x float> splat (float 1.000000e+00) to <1 x i32>)
+; CHECK-LABEL: @bitcast_constexpr_scalar_fp_to_vector_int(
+; CHECK-NEXT: ret <1 x i32> splat (i32 1065353216)
;
%res = bitcast float 1.0 to <1 x i32>
ret <1 x i32> %res
|
paulwalker-arm
changed the title
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 fixes the code quality issue reported in #111149.