Skip to content

[CIR][CIRGen][Builtin] Add several elementwise FP builtins #1553

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 1 commit into from
Apr 9, 2025
Merged

[CIR][CIRGen][Builtin] Add several elementwise FP builtins #1553

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
21 changes: 10 additions & 11 deletions clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -925,9 +925,6 @@ RValue CIRGenFunction::emitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
assert(!cir::MissingFeatures::fastMathFlags());
return emitUnaryMaybeConstrainedFPBuiltin<cir::SqrtOp>(*this, *E);

case Builtin::BI__builtin_elementwise_sqrt:
llvm_unreachable("BI__builtin_elementwise_sqrt NYI");

case Builtin::BItan:
case Builtin::BItanf:
case Builtin::BItanl:
Expand Down Expand Up @@ -1467,31 +1464,33 @@ RValue CIRGenFunction::emitBuiltinExpr(const GlobalDecl GD, unsigned BuiltinID,
case Builtin::BI__builtin_elementwise_bitreverse:
llvm_unreachable("BI__builtin_elementwise_bitreverse NYI");
case Builtin::BI__builtin_elementwise_cos:
llvm_unreachable("BI__builtin_elementwise_cos NYI");
return emitUnaryFPBuiltin<cir::CosOp>(*this, *E);
case Builtin::BI__builtin_elementwise_cosh:
llvm_unreachable("BI__builtin_elementwise_cosh NYI");
case Builtin::BI__builtin_elementwise_floor:
llvm_unreachable("BI__builtin_elementwise_floor NYI");
return emitUnaryFPBuiltin<cir::FloorOp>(*this, *E);
case Builtin::BI__builtin_elementwise_popcount:
llvm_unreachable("BI__builtin_elementwise_popcount NYI");
case Builtin::BI__builtin_elementwise_roundeven:
llvm_unreachable("BI__builtin_elementwise_roundeven NYI");
case Builtin::BI__builtin_elementwise_round:
llvm_unreachable("BI__builtin_elementwise_round NYI");
return emitUnaryFPBuiltin<cir::RoundOp>(*this, *E);
case Builtin::BI__builtin_elementwise_rint:
llvm_unreachable("BI__builtin_elementwise_rint NYI");
return emitUnaryFPBuiltin<cir::RintOp>(*this, *E);
case Builtin::BI__builtin_elementwise_nearbyint:
llvm_unreachable("BI__builtin_elementwise_nearbyint NYI");
return emitUnaryFPBuiltin<cir::NearbyintOp>(*this, *E);
case Builtin::BI__builtin_elementwise_sin:
llvm_unreachable("BI__builtin_elementwise_sin NYI");
return emitUnaryFPBuiltin<cir::SinOp>(*this, *E);
case Builtin::BI__builtin_elementwise_sinh:
llvm_unreachable("BI__builtin_elementwise_sinh NYI");
case Builtin::BI__builtin_elementwise_sqrt:
return emitUnaryFPBuiltin<cir::SqrtOp>(*this, *E);
case Builtin::BI__builtin_elementwise_tan:
llvm_unreachable("BI__builtin_elementwise_tan NYI");
return emitUnaryFPBuiltin<cir::TanOp>(*this, *E);
case Builtin::BI__builtin_elementwise_tanh:
llvm_unreachable("BI__builtin_elementwise_tanh NYI");
case Builtin::BI__builtin_elementwise_trunc:
llvm_unreachable("BI__builtin_elementwise_trunc NYI");
return emitUnaryFPBuiltin<cir::TruncOp>(*this, *E);
case Builtin::BI__builtin_elementwise_canonicalize:
llvm_unreachable("BI__builtin_elementwise_canonicalize NYI");
case Builtin::BI__builtin_elementwise_copysign:
Expand Down
189 changes: 189 additions & 0 deletions clang/test/CIR/CodeGen/builtins-elementwise.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -183,3 +183,192 @@ void test_builtin_elementwise_log10(float f, double d, vfloat4 vf4,
// LLVM: {{%.*}} = call <4 x double> @llvm.log10.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_log10(vd4);
}

void test_builtin_elementwise_cos(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_cos
// LLVM-LABEL: test_builtin_elementwise_cos
// CIR: {{%.*}} = cir.cos {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.cos.f32(float {{%.*}})
f = __builtin_elementwise_cos(f);

// CIR: {{%.*}} = cir.cos {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.cos.f64(double {{%.*}})
d = __builtin_elementwise_cos(d);

// CIR: {{%.*}} = cir.cos {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.cos.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_cos(vf4);

// CIR: {{%.*}} = cir.cos {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.cos.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_cos(vd4);
}

void test_builtin_elementwise_floor(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_floor
// LLVM-LABEL: test_builtin_elementwise_floor
// CIR: {{%.*}} = cir.floor {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.floor.f32(float {{%.*}})
f = __builtin_elementwise_floor(f);

// CIR: {{%.*}} = cir.floor {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.floor.f64(double {{%.*}})
d = __builtin_elementwise_floor(d);

// CIR: {{%.*}} = cir.floor {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.floor.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_floor(vf4);

// CIR: {{%.*}} = cir.floor {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.floor.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_floor(vd4);
}

void test_builtin_elementwise_round(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_round
// LLVM-LABEL: test_builtin_elementwise_round
// CIR: {{%.*}} = cir.round {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.round.f32(float {{%.*}})
f = __builtin_elementwise_round(f);

// CIR: {{%.*}} = cir.round {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.round.f64(double {{%.*}})
d = __builtin_elementwise_round(d);

// CIR: {{%.*}} = cir.round {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.round.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_round(vf4);

// CIR: {{%.*}} = cir.round {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.round.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_round(vd4);
}

void test_builtin_elementwise_rint(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_rint
// LLVM-LABEL: test_builtin_elementwise_rint
// CIR: {{%.*}} = cir.rint {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.rint.f32(float {{%.*}})
f = __builtin_elementwise_rint(f);

// CIR: {{%.*}} = cir.rint {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.rint.f64(double {{%.*}})
d = __builtin_elementwise_rint(d);

// CIR: {{%.*}} = cir.rint {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.rint.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_rint(vf4);

// CIR: {{%.*}} = cir.rint {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.rint.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_rint(vd4);
}

void test_builtin_elementwise_nearbyint(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_nearbyint
// LLVM-LABEL: test_builtin_elementwise_nearbyint
// CIR: {{%.*}} = cir.nearbyint {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.nearbyint.f32(float {{%.*}})
f = __builtin_elementwise_nearbyint(f);

// CIR: {{%.*}} = cir.nearbyint {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.nearbyint.f64(double {{%.*}})
d = __builtin_elementwise_nearbyint(d);

// CIR: {{%.*}} = cir.nearbyint {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_nearbyint(vf4);

// CIR: {{%.*}} = cir.nearbyint {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_nearbyint(vd4);
}

void test_builtin_elementwise_sin(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_sin
// LLVM-LABEL: test_builtin_elementwise_sin
// CIR: {{%.*}} = cir.sin {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.sin.f32(float {{%.*}})
f = __builtin_elementwise_sin(f);

// CIR: {{%.*}} = cir.sin {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.sin.f64(double {{%.*}})
d = __builtin_elementwise_sin(d);

// CIR: {{%.*}} = cir.sin {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.sin.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_sin(vf4);

// CIR: {{%.*}} = cir.sin {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.sin.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_sin(vd4);
}

void test_builtin_elementwise_sqrt(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_sqrt
// LLVM-LABEL: test_builtin_elementwise_sqrt
// CIR: {{%.*}} = cir.sqrt {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.sqrt.f32(float {{%.*}})
f = __builtin_elementwise_sqrt(f);

// CIR: {{%.*}} = cir.sqrt {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.sqrt.f64(double {{%.*}})
d = __builtin_elementwise_sqrt(d);

// CIR: {{%.*}} = cir.sqrt {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.sqrt.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_sqrt(vf4);

// CIR: {{%.*}} = cir.sqrt {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.sqrt.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_sqrt(vd4);
}

void test_builtin_elementwise_tan(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_tan
// LLVM-LABEL: test_builtin_elementwise_tan
// CIR: {{%.*}} = cir.tan {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.tan.f32(float {{%.*}})
f = __builtin_elementwise_tan(f);

// CIR: {{%.*}} = cir.tan {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.tan.f64(double {{%.*}})
d = __builtin_elementwise_tan(d);

// CIR: {{%.*}} = cir.tan {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.tan.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_tan(vf4);

// CIR: {{%.*}} = cir.tan {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.tan.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_tan(vd4);
}

void test_builtin_elementwise_trunc(float f, double d, vfloat4 vf4,
vdouble4 vd4) {
// CIR-LABEL: test_builtin_elementwise_trunc
// LLVM-LABEL: test_builtin_elementwise_trunc
// CIR: {{%.*}} = cir.trunc {{%.*}} : !cir.float
// LLVM: {{%.*}} = call float @llvm.trunc.f32(float {{%.*}})
f = __builtin_elementwise_trunc(f);

// CIR: {{%.*}} = cir.trunc {{%.*}} : !cir.double
// LLVM: {{%.*}} = call double @llvm.trunc.f64(double {{%.*}})
d = __builtin_elementwise_trunc(d);

// CIR: {{%.*}} = cir.trunc {{%.*}} : !cir.vector<!cir.float x 4>
// LLVM: {{%.*}} = call <4 x float> @llvm.trunc.v4f32(<4 x float> {{%.*}})
vf4 = __builtin_elementwise_trunc(vf4);

// CIR: {{%.*}} = cir.trunc {{%.*}} : !cir.vector<!cir.double x 4>
// LLVM: {{%.*}} = call <4 x double> @llvm.trunc.v4f64(<4 x double> {{%.*}})
vd4 = __builtin_elementwise_trunc(vd4);
}
Loading