diff options
| author | Craig Topper <craig.topper@intel.com> | 2020-02-07 15:04:22 -0800 |
|---|---|---|
| committer | Craig Topper <craig.topper@intel.com> | 2020-02-07 15:27:05 -0800 |
| commit | bb717d3f469b997ce6b13f2b87ffd78d2d2254b7 (patch) | |
| tree | a29cba1b3cd3b34ae3150aad1579c70218573bf6 | |
| parent | 598d9dd846f1086abfc1c5055df0d25e8a485fc9 (diff) | |
| download | llvm-bb717d3f469b997ce6b13f2b87ffd78d2d2254b7.zip llvm-bb717d3f469b997ce6b13f2b87ffd78d2d2254b7.tar.gz llvm-bb717d3f469b997ce6b13f2b87ffd78d2d2254b7.tar.bz2 | |
[X86] Correct the implementation of the avx512 masked fmsubadd autoupgrade code to not leave the negate unconnected.
This was causing us to generate fmaddsub instead of fmsubadd if
rounding control is not 4.
| -rw-r--r-- | llvm/lib/IR/AutoUpgrade.cpp | 3 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll | 4 |
2 files changed, 3 insertions, 4 deletions
diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp index a14bcef9..faea5b0 100644 --- a/llvm/lib/IR/AutoUpgrade.cpp +++ b/llvm/lib/IR/AutoUpgrade.cpp @@ -3259,8 +3259,7 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) { Ops[2] = Builder.CreateFNeg(Ops[2]); Rep = Builder.CreateCall(Intrinsic::getDeclaration(F->getParent(), IID), - {CI->getArgOperand(0), CI->getArgOperand(1), - CI->getArgOperand(2), CI->getArgOperand(4)}); + Ops); } else { int NumElts = CI->getType()->getVectorNumElements(); diff --git a/llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll index 39746ff..0ab6091 100644 --- a/llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll +++ b/llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll @@ -735,7 +735,7 @@ define <8 x double>@test_int_x86_avx512_mask3_vfmsubadd_pd_512_rne(<8 x double> ; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_512_rne: ; CHECK: ## %bb.0: ; CHECK-NEXT: kmovw %edi, %k1 -; CHECK-NEXT: vfmaddsub231pd {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1} +; CHECK-NEXT: vfmsubadd231pd {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1} ; CHECK-NEXT: vmovapd %zmm2, %zmm0 ; CHECK-NEXT: retq %res = call <8 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 8) @@ -746,7 +746,7 @@ define <16 x float>@test_int_x86_avx512_mask3_vfmsubadd_ps_512_rne(<16 x float> ; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_512_rne: ; CHECK: ## %bb.0: ; CHECK-NEXT: kmovw %edi, %k1 -; CHECK-NEXT: vfmaddsub231ps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1} +; CHECK-NEXT: vfmsubadd231ps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1} ; CHECK-NEXT: vmovaps %zmm2, %zmm0 ; CHECK-NEXT: retq %res = call <16 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 8) |