[X86][AVX] SimplifyDemandedVectorEltsForTargetShuffle - ensure mask is same size as constant size

Fixes test regression reported on D81791
This commit is contained in:
Simon Pilgrim 2020-07-08 11:47:43 +01:00
parent 1be92dd207
commit 9dc250db9d
2 changed files with 13 additions and 6 deletions

View File

@ -36935,11 +36935,16 @@ bool X86TargetLowering::SimplifyDemandedVectorEltsForTargetShuffle(
return false;
const Constant *C = getTargetConstantFromNode(Load);
if (!C || !C->getType()->isVectorTy())
if (!C)
return false;
Type *CTy = C->getType();
if (!CTy->isVectorTy() ||
CTy->getPrimitiveSizeInBits() != Mask.getValueSizeInBits())
return false;
// Handle scaling for i64 elements on 32-bit targets.
unsigned NumCstElts = cast<FixedVectorType>(C->getType())->getNumElements();
unsigned NumCstElts = cast<FixedVectorType>(CTy)->getNumElements();
if (NumCstElts != NumElts && NumCstElts != (NumElts * 2))
return false;
unsigned Scale = NumCstElts / NumElts;

View File

@ -528,8 +528,9 @@ define void @test_demandedelts_pshufb_v32i8_v16i8(<2 x i32>* %src, <8 x i32>* %d
; SKX64-NEXT: vmovdqa 32(%rdi), %xmm0
; SKX64-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[12,13,14,15,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
; SKX64-NEXT: vmovdqa %ymm0, 672(%rsi)
; SKX64-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,0,2,3]
; SKX64-NEXT: vmovaps %ymm0, 832(%rsi)
; SKX64-NEXT: vmovdqa 208(%rdi), %xmm0
; SKX64-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,6,7,0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero
; SKX64-NEXT: vmovdqa %ymm0, 832(%rsi)
; SKX64-NEXT: vzeroupper
; SKX64-NEXT: retq
;
@ -550,8 +551,9 @@ define void @test_demandedelts_pshufb_v32i8_v16i8(<2 x i32>* %src, <8 x i32>* %d
; SKX32-NEXT: vmovdqa 32(%ecx), %xmm0
; SKX32-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[12,13,14,15,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
; SKX32-NEXT: vmovdqa %ymm0, 672(%eax)
; SKX32-NEXT: vpermilps {{.*#+}} xmm0 = mem[1,0,2,3]
; SKX32-NEXT: vmovaps %ymm0, 832(%eax)
; SKX32-NEXT: vmovdqa 208(%ecx), %xmm0
; SKX32-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,6,7,0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero
; SKX32-NEXT: vmovdqa %ymm0, 832(%eax)
; SKX32-NEXT: vzeroupper
; SKX32-NEXT: retl
;