RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-12-13 14:47:00 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[X86][SSE] When lowering a 256-bit shuffle as PMOVZX, reduce the input vector to the lower 128-bit subvector.

Browse Source 
Most often as not this is what it started out as, the extraction is zero-cost on AVX and the PMOVZX/PMOVSX folding logic is based around 128-bit loads.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@270858 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Simon Pilgrim 2016-05-26 15:40:36 +00:00
parent 32afd6e7e7
commit 1f433c478c
2 changed files with 9 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
lib/Target/X86/X86ISelLowering.cpp
View File

@ -7915,7 +7915,13 @@ static SDValue lowerVectorShuffleAsSpecificZeroOrAnyExtend(
return SDValue();
MVT ExtVT = MVT::getVectorVT(MVT::getIntegerVT(EltBits * Scale),
NumElements / Scale);
InputV = DAG.getNode(X86ISD::VZEXT, DL, ExtVT, ShuffleOffset(InputV));
InputV = ShuffleOffset(InputV);
// For 256-bit vectors, we only need the lower (128-bit) input half.
if (VT.is256BitVector())
InputV = extract128BitVector(InputV, 0, DAG, DL);
InputV = DAG.getNode(X86ISD::VZEXT, DL, ExtVT, InputV);
return DAG.getBitcast(VT, InputV);
}

6
test/CodeGen/X86/vector-zext.ll
View File

@ -707,14 +707,12 @@ define <8 x i32> @load_zext_16i8_to_8i32(<16 x i8> *%ptr) {
;
; AVX2-LABEL: load_zext_16i8_to_8i32:
; AVX2: # BB#0: # %entry
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; AVX2-NEXT: vpmovzxbd {{.*#+}} ymm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
; AVX2-NEXT: retq
;
; AVX512-LABEL: load_zext_16i8_to_8i32:
; AVX512: # BB#0: # %entry
; AVX512-NEXT: vmovdqa (%rdi), %xmm0
; AVX512-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; AVX512-NEXT: vpmovzxbd {{.*#+}} ymm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
; AVX512-NEXT: retq
entry:
%X = load <16 x i8>, <16 x i8>* %ptr