r196361 - [AArch64 NEON] Add ACLE intrinsic vceqz_f64.
Kevin Qin
Kevin.Qin at arm.com
Wed Dec 4 00:02:11 PST 2013
Author: kevinqin
Date: Wed Dec 4 02:02:11 2013
New Revision: 196361
URL: http://llvm.org/viewvc/llvm-project?rev=196361&view=rev
Log:
[AArch64 NEON] Add ACLE intrinsic vceqz_f64.
Modified:
cfe/trunk/lib/CodeGen/CGBuiltin.cpp
cfe/trunk/test/CodeGen/aarch64-neon-misc.c
Modified: cfe/trunk/lib/CodeGen/CGBuiltin.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGBuiltin.cpp?rev=196361&r1=196360&r2=196361&view=diff
==============================================================================
--- cfe/trunk/lib/CodeGen/CGBuiltin.cpp (original)
+++ cfe/trunk/lib/CodeGen/CGBuiltin.cpp Wed Dec 4 02:02:11 2013
@@ -1759,6 +1759,7 @@ static Value *EmitAArch64ScalarBuiltinEx
bool ExtendEle = false;
bool OverloadInt = false;
bool OverloadCmpInt = false;
+ bool IsFpCmpZInt = false;
bool OverloadCvtInt = false;
bool OverloadWideInt = false;
bool OverloadNarrowInt = false;
@@ -2269,7 +2270,8 @@ static Value *EmitAArch64ScalarBuiltinEx
case AArch64::BI__builtin_neon_vceqzd_f64:
Int = Intrinsic::aarch64_neon_vceq; s = "vceq";
// Add implicit zero operand.
- Ops.push_back(llvm::Constant::getNullValue(Ops[0]->getType()));
+ Ops.push_back(llvm::Constant::getNullValue(CGF.FloatTy));
+ IsFpCmpZInt = true;
OverloadCmpInt = true; break;
// Scalar Floating-point Compare Greater Than Or Equal
case AArch64::BI__builtin_neon_vcges_f32:
@@ -2281,7 +2283,8 @@ static Value *EmitAArch64ScalarBuiltinEx
case AArch64::BI__builtin_neon_vcgezd_f64:
Int = Intrinsic::aarch64_neon_vcge; s = "vcge";
// Add implicit zero operand.
- Ops.push_back(llvm::Constant::getNullValue(Ops[0]->getType()));
+ Ops.push_back(llvm::Constant::getNullValue(CGF.FloatTy));
+ IsFpCmpZInt = true;
OverloadCmpInt = true; break;
// Scalar Floating-point Compare Greather Than
case AArch64::BI__builtin_neon_vcgts_f32:
@@ -2293,7 +2296,8 @@ static Value *EmitAArch64ScalarBuiltinEx
case AArch64::BI__builtin_neon_vcgtzd_f64:
Int = Intrinsic::aarch64_neon_vcgt; s = "vcgt";
// Add implicit zero operand.
- Ops.push_back(llvm::Constant::getNullValue(Ops[0]->getType()));
+ Ops.push_back(llvm::Constant::getNullValue(CGF.FloatTy));
+ IsFpCmpZInt = true;
OverloadCmpInt = true; break;
// Scalar Floating-point Compare Less Than or Equal
case AArch64::BI__builtin_neon_vcles_f32:
@@ -2305,7 +2309,8 @@ static Value *EmitAArch64ScalarBuiltinEx
case AArch64::BI__builtin_neon_vclezd_f64:
Int = Intrinsic::aarch64_neon_vclez; s = "vcle";
// Add implicit zero operand.
- Ops.push_back(llvm::Constant::getNullValue(Ops[0]->getType()));
+ Ops.push_back(llvm::Constant::getNullValue(CGF.FloatTy));
+ IsFpCmpZInt = true;
OverloadCmpInt = true; break;
// Scalar Floating-point Compare Less Than Zero
case AArch64::BI__builtin_neon_vclts_f32:
@@ -2317,7 +2322,8 @@ static Value *EmitAArch64ScalarBuiltinEx
case AArch64::BI__builtin_neon_vcltzd_f64:
Int = Intrinsic::aarch64_neon_vcltz; s = "vclt";
// Add implicit zero operand.
- Ops.push_back(llvm::Constant::getNullValue(Ops[0]->getType()));
+ Ops.push_back(llvm::Constant::getNullValue(CGF.FloatTy));
+ IsFpCmpZInt = true;
OverloadCmpInt = true; break;
// Scalar Floating-point Absolute Compare Greater Than Or Equal
case AArch64::BI__builtin_neon_vcages_f32:
@@ -2601,6 +2607,8 @@ static Value *EmitAArch64ScalarBuiltinEx
Ty = CGF.ConvertType(Arg->getType());
VTy = llvm::VectorType::get(Ty, 1);
Tys.push_back(VTy);
+ if(IsFpCmpZInt)
+ VTy = llvm::VectorType::get(CGF.FloatTy, 1);
Tys.push_back(VTy);
F = CGF.CGM.getIntrinsic(Int, Tys);
Modified: cfe/trunk/test/CodeGen/aarch64-neon-misc.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/aarch64-neon-misc.c?rev=196361&r1=196360&r2=196361&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/aarch64-neon-misc.c (original)
+++ cfe/trunk/test/CodeGen/aarch64-neon-misc.c Wed Dec 4 02:02:11 2013
@@ -114,6 +114,12 @@ uint32x2_t test_vceqz_f32(float32x2_t a)
return vceqz_f32(a);
}
+// CHECK: test_vceqz_f64
+// CHECK: fcmeq {{d[0-9]+}}, {{d[0-9]+}}, #0
+uint64x1_t test_vceqz_f64(float64x1_t a) {
+ return vceqz_f64(a);
+}
+
// CHECK: test_vceqzq_f32
// CHECK: fcmeq {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, #0
uint32x4_t test_vceqzq_f32(float32x4_t a) {
More information about the cfe-commits
mailing list