shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

[clang][CodeGen] Propagate pragma set fast-math flags to floating point builtins (#90377)

Browse Source 
This is a fix for the issue #87758 where fast-math flags are not
propagated all builtins.

It seems like pragmas with fast math flags was only propagated to calls
of unary floating point builtins. This patch propagate them also for
binary and ternary floating point builtins.
This commit is contained in:
Karl-Johan Karlsson 2024-05-04 17:47:48 +02:00 committed by GitHub
parent 7ee6288312
commit cb015b9ec9
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 84 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
clang/lib/CodeGen/CGBuiltin.cpp
View File

@ -513,8 +513,8 @@ static Value *emitBinaryMaybeConstrainedFPBuiltin(CodeGenFunction &CGF,
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
if (CGF.Builder.getIsFPConstrained()) {
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
Function *F = CGF.CGM.getIntrinsic(ConstrainedIntrinsicID, Src0->getType());
return CGF.Builder.CreateConstrainedFPCall(F, { Src0, Src1 });
} else {
@ -530,8 +530,8 @@ static Value *emitBinaryExpMaybeConstrainedFPBuiltin(
llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
if (CGF.Builder.getIsFPConstrained()) {
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
Function *F = CGF.CGM.getIntrinsic(ConstrainedIntrinsicID,
{Src0->getType(), Src1->getType()});
return CGF.Builder.CreateConstrainedFPCall(F, {Src0, Src1});
@ -551,8 +551,8 @@ static Value *emitTernaryMaybeConstrainedFPBuiltin(CodeGenFunction &CGF,
llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
llvm::Value *Src2 = CGF.EmitScalarExpr(E->getArg(2));
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
if (CGF.Builder.getIsFPConstrained()) {
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
Function *F = CGF.CGM.getIntrinsic(ConstrainedIntrinsicID, Src0->getType());
return CGF.Builder.CreateConstrainedFPCall(F, { Src0, Src1, Src2 });
} else {
@ -704,6 +704,7 @@ static Value *EmitSignBit(CodeGenFunction &CGF, Value *V) {
static RValue emitLibraryCall(CodeGenFunction &CGF, const FunctionDecl *FD,
const CallExpr *E, llvm::Constant *calleeValue) {
CodeGenFunction::CGFPOptionsRAII FPOptsRAII(CGF, E);
CGCallee callee = CGCallee::forDirect(calleeValue, GlobalDecl(FD));
return CGF.EmitCall(E->getCallee()->getType(), callee, E, ReturnValueSlot());
}
@ -2660,7 +2661,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
if (OP.hasMathErrnoOverride())
ErrnoOverriden = OP.getMathErrnoOverride();
}
// True if 'atttibute__((optnone)) is used. This attibute overrides
// True if 'attribute__((optnone))' is used. This attribute overrides
// fast-math which implies math-errno.
bool OptNone = CurFuncDecl && CurFuncDecl->hasAttr<OptimizeNoneAttr>();

6
clang/test/CodeGen/math-errno.c
View File

@ -27,7 +27,7 @@ float f1(float x) {
// CHECK: tail call float @sqrtf(float noundef {{.*}}) #[[ATTR4_O2:[0-9]+]]
// FAST-LABEL: define {{.*}} nofpclass(nan inf) float @f1
// FAST: call fast nofpclass(nan inf) float @sqrtf(float noundef nofpclass(nan inf) {{.*}}) #[[ATTR3_FAST:[0-9]+]]
// FAST: call nofpclass(nan inf) float @sqrtf(float noundef nofpclass(nan inf) {{.*}}) #[[ATTR3_FAST:[0-9]+]]
// NOOPT-LABEL: define {{.*}} float @f1
// NOOPT: call float @sqrtf(float noundef {{.*}}) #[[ATTR4_NOOPT:[0-9]+]]
@ -44,7 +44,7 @@ float f2(float x) {
// FAST: call fast float @llvm.sqrt.f32(float {{.*}})
// NOOPT-LABEL: define {{.*}} float @f2
// NOOPT: call float @sqrtf(float {{.*}}) #[[ATTR4_NOOPT:[0-9]+]]
// NOOPT: call fast float @sqrtf(float {{.*}}) #[[ATTR4_NOOPT:[0-9]+]]
__attribute__((optnone))
float f3(float x) {
@ -56,7 +56,7 @@ float f3(float x) {
// CHECK: call float @sqrtf(float noundef {{.*}})
// FAST-LABEL: define {{.*}} nofpclass(nan inf) float @f3
// FAST: call fast nofpclass(nan inf) float @sqrtf(float noundef nofpclass(nan inf) {{.*}}) #[[ATTR4_FAST:[0-9]+]]
// FAST: call nofpclass(nan inf) float @sqrtf(float noundef nofpclass(nan inf) {{.*}}) #[[ATTR4_FAST:[0-9]+]]
// NOOPT-LABEL: define {{.*}} float @f3
// NOOPT: call float @sqrtf(float noundef %0) #[[ATTR4_NOOPT:[0-9]+]]

76
clang/test/CodeGen/pr87758.c Normal file
View File

@ -0,0 +1,76 @@
// RUN: %clang_cc1 -triple x86_64-linux-gnu -emit-llvm -o - %s | FileCheck %s
// precise mode
// RUN: %clang_cc1 -triple x86_64-linux-gnu -fmath-errno -ffp-contract=on \
// RUN: -fno-rounding-math -emit-llvm -o - %s | FileCheck \
// RUN: --check-prefix=CHECK-PRECISE %s
// fast mode
// RUN: %clang_cc1 -triple x86_64-linux-gnu -ffast-math -ffp-contract=fast \
// RUN: -emit-llvm -o - %s | FileCheck --check-prefix=CHECK-FAST %s
// Reproducer for issue #87758
// The testcase below verifies that the "fast" flag are set on the calls.
float sqrtf(float x); // unary fp builtin
float powf(float x, float y); // binary fp builtin
float fmaf(float x, float y, float z); // ternary fp builtin
char *rindex(const char *s, int c); // not a fp builtin
#pragma float_control(push)
#pragma float_control(precise, off)
// CHECK: define dso_local float @fp_precise_off_libm_calls(
// CHECK: call fast float @llvm.sqrt.f32(
// CHECK: call fast float @llvm.pow.f32(
// CHECK: call fast float @llvm.fma.f32(
// CHECK: call ptr @rindex(
// CHECK-PRECISE: define dso_local float @fp_precise_off_libm_calls(
// CHECK-PRECISE: call fast float @sqrtf(
// CHECK-PRECISE: call fast float @powf(
// CHECK-PRECISE: call fast float @llvm.fma.f32(
// CHECK-PRECISE: call ptr @rindex(
// CHECK-FAST: define dso_local nofpclass(nan inf) float @fp_precise_off_libm_calls(
// CHECK-FAST: call fast float @llvm.sqrt.f32(
// CHECK-FAST: call fast float @llvm.pow.f32(
// CHECK-FAST: call fast float @llvm.fma.f32(
// CHECK-FAST: call ptr @rindex(
float fp_precise_off_libm_calls(float a, float b, float c, const char *d, char *e, unsigned char f) {
a = sqrtf(a);
a = powf(a,b);
a = fmaf(a,b,c);
e = rindex(d, 75);
return a;
}
#pragma float_control(pop)
#pragma float_control(push)
#pragma float_control(precise, on)
// CHECK: define dso_local float @fp_precise_on_libm_calls(
// CHECK: call float @sqrtf(
// CHECK: call float @powf(
// CHECK: call float @llvm.fma.f32(
// CHECK: call ptr @rindex(
// CHECK-PRECISE: define dso_local float @fp_precise_on_libm_calls(
// CHECK-PRECISE: call float @sqrtf(
// CHECK-PRECISE: call float @powf(
// CHECK-PRECISE: call float @llvm.fma.f32(
// CHECK-PRECISE: call ptr @rindex(
// CHECK-FAST: define dso_local nofpclass(nan inf) float @fp_precise_on_libm_calls(
// CHECK-FAST: call nofpclass(nan inf) float @sqrtf(
// CHECK-FAST: call nofpclass(nan inf) float @powf(
// CHECK-FAST: call float @llvm.fma.f32(
// CHECK-FAST: call ptr @rindex(
float fp_precise_on_libm_calls(float a, float b, float c, const char *d, char *e, unsigned char f) {
a = sqrtf(a);
a = powf(a,b);
a = fmaf(a,b,c);
e = rindex(d, 75);
return a;
}
#pragma float_control(pop)