[AArch64] Use fneg instead of fsub -0.0, X Cin IR expansion of __builtin_neon_vfmsh_f16.

Addresses the FIXME and removes the only in tree use of
llvm::ConstantFP::getZeroValueForNegation for an FP type.

Reviewed By: dmgreen, SjoerdMeijer

Differential Revision: https://reviews.llvm.org/D147497

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index f399b07..1d79d9a 100644 (file)
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -10965,14 +10965,12 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
         *this, Intrinsic::fma, Intrinsic::experimental_constrained_fma, HalfTy,
         {EmitScalarExpr(E->getArg(1)), EmitScalarExpr(E->getArg(2)), Ops[0]});
   case NEON::BI__builtin_neon_vfmsh_f16: {
-    // FIXME: This should be an fneg instruction:
-    Value *Zero = llvm::ConstantFP::getZeroValueForNegation(HalfTy);
-    Value* Sub = Builder.CreateFSub(Zero, EmitScalarExpr(E->getArg(1)), "vsubh");
+    Value* Neg = Builder.CreateFNeg(EmitScalarExpr(E->getArg(1)), "vsubh");
 
     // NEON intrinsic puts accumulator first, unlike the LLVM fma.
     return emitCallMaybeConstrainedFPBuiltin(
         *this, Intrinsic::fma, Intrinsic::experimental_constrained_fma, HalfTy,
-        {Sub, EmitScalarExpr(E->getArg(2)), Ops[0]});
+        {Neg, EmitScalarExpr(E->getArg(2)), Ops[0]});
   }
   case NEON::BI__builtin_neon_vaddd_s64:
   case NEON::BI__builtin_neon_vaddd_u64:

diff --git a/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c b/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
index 198c452..5367134 100644 (file)
--- a/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
+++ b/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
@@ -290,8 +290,7 @@ float16_t test_vfmah_f16(float16_t a, float16_t b, float16_t c) {
 }
 
 // COMMON-LABEL: test_vfmsh_f16
-// UNCONSTRAINED:  [[SUB:%.*]] = fsub half 0xH8000, %b
-// CONSTRAINED:    [[SUB:%.*]] = call half @llvm.experimental.constrained.fsub.f16(half 0xH8000, half %b, metadata !"round.tonearest", metadata !"fpexcept.strict")
+// COMMONIR:  [[SUB:%.*]] = fneg half %b
 // UNCONSTRAINED:  [[ADD:%.*]] = call half @llvm.fma.f16(half [[SUB]], half %c, half %a)
 // CONSTRAINED:    [[ADD:%.*]] = call half @llvm.experimental.constrained.fma.f16(half [[SUB]], half %c, half %a, metadata !"round.tonearest", metadata !"fpexcept.strict")
 // COMMONIR:       ret half [[ADD]]

diff --git a/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c b/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
index b374ba1..d745a77 100644 (file)
--- a/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
+++ b/clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
@@ -652,7 +652,7 @@ float16_t test_vfmah_f16(float16_t a, float16_t b, float16_t c) {
 }
 
 // CHECK-LABEL: test_vfmsh_f16
-// CHECK:  [[SUB:%.*]] = fsub half 0xH8000, %b
+// CHECK:  [[SUB:%.*]] = fneg half %b
 // CHECK:  [[ADD:%.*]] = call half @llvm.fma.f16(half [[SUB]], half %c, half %a)
 // CHECK:  ret half [[ADD]]
 float16_t test_vfmsh_f16(float16_t a, float16_t b, float16_t c) {