RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 18:54:02 +01:00
Code Issues Projects Releases Wiki Activity

[X86] More aggressive shuffle mask widening in combineExtractWithShuffle

Browse Source 
Use demanded extract index to set most of the shuffle mask to undef, making it easier to widen and peek through.

llvm-svn: 351013
This commit is contained in:
Simon Pilgrim 2019-01-12 16:38:56 +00:00
parent fc97f679c1
commit a89e77f814
2 changed files with 10 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
lib/Target/X86/X86ISelLowering.cpp
View File

@ -33691,6 +33691,15 @@ static SDValue combineExtractWithShuffle(SDNode *N, SelectionDAG &DAG,
scaleShuffleMask<int>(Scale, Mask, ScaledMask);
Mask = std::move(ScaledMask);
} else if ((Mask.size() % NumSrcElts) == 0) {
// Simplify Mask based on demanded element.
int ExtractIdx = (int)N->getConstantOperandVal(1);
int Scale = Mask.size() / NumSrcElts;
int Lo = Scale * ExtractIdx;
int Hi = Scale * (ExtractIdx + 1);
for (int i = 0, e = (int)Mask.size(); i != e; ++i)
if (i < Lo || Hi <= i)
Mask[i] = SM_SentinelUndef;
SmallVector<int, 16> WidenedMask;
while (Mask.size() > NumSrcElts &&
canWidenShuffleElements(Mask, WidenedMask))

3
test/CodeGen/X86/and-load-fold.ll
View File

@ -7,8 +7,7 @@ define i8 @foo(<4 x i8>* %V) {
; CHECK-LABEL: foo:
; CHECK: # %bb.0:
; CHECK-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; CHECK-NEXT: pextrw $2, %xmm0, %eax
; CHECK-NEXT: pextrw $1, %xmm0, %eax
; CHECK-NEXT: andb $95, %al
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: retq