This revision was landed with ongoing or failed builds.
This revision was automatically updated to reflect the committed changes.
Closed by commit rG0109f8d1e3bf: [AArch64] Use fneg instead of fsub -0.0, X Cin
IR expansion of… (authored by craig.topper).
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D147497/new/
https://reviews.llvm.org/D147497
Files:
clang/lib/CodeGen/CGBuiltin.cpp
clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
Index: clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
===================================================================
--- clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
+++ clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
@@ -652,7 +652,7 @@
}
// 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) {
Index: clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
===================================================================
--- clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
+++ clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
@@ -290,8 +290,7 @@
}
// 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]]
Index: clang/lib/CodeGen/CGBuiltin.cpp
===================================================================
--- clang/lib/CodeGen/CGBuiltin.cpp
+++ clang/lib/CodeGen/CGBuiltin.cpp
@@ -10965,14 +10965,12 @@
*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:
Index: clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
===================================================================
--- clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
+++ clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c
@@ -652,7 +652,7 @@
}
// 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) {
Index: clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
===================================================================
--- clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
+++ clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics-constrained.c
@@ -290,8 +290,7 @@
}
// 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]]
Index: clang/lib/CodeGen/CGBuiltin.cpp
===================================================================
--- clang/lib/CodeGen/CGBuiltin.cpp
+++ clang/lib/CodeGen/CGBuiltin.cpp
@@ -10965,14 +10965,12 @@
*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:
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
