diff options
| -rw-r--r-- | llvm/lib/Target/X86/X86ISelLowering.cpp | 14 | ||||
| -rw-r--r-- | llvm/test/CodeGen/X86/shuffle-vs-trunc-256-widen.ll | 4 |
2 files changed, 16 insertions, 2 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 7a2e842c2b9..d881615c8a2 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -33039,6 +33039,20 @@ bool X86TargetLowering::SimplifyDemandedVectorEltsForTargetNode( return true; break; } + case X86ISD::VTRUNC: + case X86ISD::VTRUNCS: + case X86ISD::VTRUNCUS: { + SDValue Src = Op.getOperand(0); + MVT SrcVT = Src.getSimpleValueType(); + APInt DemandedSrc = DemandedElts.zextOrTrunc(SrcVT.getVectorNumElements()); + APInt SrcUndef, SrcZero; + if (SimplifyDemandedVectorElts(Src, DemandedSrc, SrcUndef, SrcZero, TLO, + Depth + 1)) + return true; + KnownZero = SrcZero.zextOrTrunc(NumElts); + KnownUndef = SrcUndef.zextOrTrunc(NumElts); + break; + } case X86ISD::BLENDV: { APInt SelUndef, SelZero; if (SimplifyDemandedVectorElts(Op.getOperand(0), DemandedElts, SelUndef, diff --git a/llvm/test/CodeGen/X86/shuffle-vs-trunc-256-widen.ll b/llvm/test/CodeGen/X86/shuffle-vs-trunc-256-widen.ll index e2e4b43c8ce..1f2adabe5fa 100644 --- a/llvm/test/CodeGen/X86/shuffle-vs-trunc-256-widen.ll +++ b/llvm/test/CodeGen/X86/shuffle-vs-trunc-256-widen.ll @@ -1179,7 +1179,7 @@ define void @trunc_v4i64_to_v4i8(<32 x i8>* %L, <4 x i8>* %S) nounwind { ; ; AVX512F-LABEL: trunc_v4i64_to_v4i8: ; AVX512F: # %bb.0: -; AVX512F-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = mem[0,1,2,3,0,1,2,3] +; AVX512F-NEXT: vmovdqa (%rdi), %ymm0 ; AVX512F-NEXT: vpmovqb %zmm0, %xmm0 ; AVX512F-NEXT: vmovd %xmm0, (%rsi) ; AVX512F-NEXT: vzeroupper @@ -1194,7 +1194,7 @@ define void @trunc_v4i64_to_v4i8(<32 x i8>* %L, <4 x i8>* %S) nounwind { ; ; AVX512BW-LABEL: trunc_v4i64_to_v4i8: ; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vbroadcasti64x4 {{.*#+}} zmm0 = mem[0,1,2,3,0,1,2,3] +; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0 ; AVX512BW-NEXT: vpmovqb %zmm0, %xmm0 ; AVX512BW-NEXT: vmovd %xmm0, (%rsi) ; AVX512BW-NEXT: vzeroupper |
