RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 04:32:44 +01:00
Code Issues Projects Releases Wiki Activity

[X86] SimplifyDemandedVectorElts - add X86ISD::VPERMV3 mask support

Browse Source 
Completes SimplifyDemandedVectorElts's basic variable shuffle mask support which should help D60512 + D60562 

llvm-svn: 358186
This commit is contained in:
Simon Pilgrim 2019-04-11 15:29:15 +00:00
parent c3542e3376
commit 6d8eb7f280
2 changed files with 3 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Target/X86/X86ISelLowering.cpp
View File

@ -33234,8 +33234,8 @@ bool X86TargetLowering::SimplifyDemandedVectorEltsForTargetNode(
break;
}
case X86ISD::PSHUFB:
case X86ISD::VPERMV3:
case X86ISD::VPERMILPV: {
// TODO - simplify other variable shuffle masks.
SDValue Mask = Op.getOperand(1);
APInt MaskUndef, MaskZero;
if (SimplifyDemandedVectorElts(Mask, DemandedElts, MaskUndef, MaskZero, TLO,

6
test/CodeGen/X86/vector-shuffle-combining-avx512bw.ll
View File

@ -933,10 +933,8 @@ define <8 x double> @combine_vpermi2var_8f64_as_permpd(<8 x double> %x0, <8 x do
;
; X64-LABEL: combine_vpermi2var_8f64_as_permpd:
; X64: # %bb.0:
; X64-NEXT: vmovdqa {{.*#+}} xmm2 = <u,2,1,3,4,6,5,7>
; X64-NEXT: vpinsrq $0, %rdi, %xmm2, %xmm2
; X64-NEXT: vmovdqa64 {{.*#+}} zmm3 = <u,2,1,3,4,6,5,7>
; X64-NEXT: vinserti32x4 $0, %xmm2, %zmm3, %zmm2
; X64-NEXT: vmovapd {{.*#+}} zmm2 = <u,2,1,3,4,6,5,7>
; X64-NEXT: vinsertf32x4 $0, {{.*}}(%rip), %zmm2, %zmm2
; X64-NEXT: vpermi2pd %zmm1, %zmm0, %zmm2
; X64-NEXT: vpermpd {{.*#+}} zmm0 = zmm2[2,3,1,1,6,7,5,5]
; X64-NEXT: retq