[llvm] r343919 - [X86][SSE] SimplifyDemandedVectorEltsForTargetNode - simplify PSHUFB masks
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sat Oct 6 06:49:31 PDT 2018
Author: rksimon
Date: Sat Oct 6 06:49:31 2018
New Revision: 343919
URL: http://llvm.org/viewvc/llvm-project?rev=343919&view=rev
Log:
[X86][SSE] SimplifyDemandedVectorEltsForTargetNode - simplify PSHUFB masks
Attempt to simplify PSHUFB masks (even non-constant ones) - we should probably be able to simplify other variable shuffles as well as the need arises.
Modified:
llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
llvm/trunk/test/CodeGen/X86/oddshuffles.ll
llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v8.ll
llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll
Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=343919&r1=343918&r2=343919&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Sat Oct 6 06:49:31 2018
@@ -31755,6 +31755,15 @@ bool X86TargetLowering::SimplifyDemanded
return true;
break;
}
+ case X86ISD::PSHUFB: {
+ // TODO - simplify other variable shuffle masks.
+ SDValue Mask = Op.getOperand(1);
+ APInt MaskUndef, MaskZero;
+ if (SimplifyDemandedVectorElts(Mask, DemandedElts, MaskUndef, MaskZero, TLO,
+ Depth + 1))
+ return true;
+ break;
+ }
}
// Simplify target shuffles.
Modified: llvm/trunk/test/CodeGen/X86/oddshuffles.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/oddshuffles.ll?rev=343919&r1=343918&r2=343919&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/oddshuffles.ll (original)
+++ llvm/trunk/test/CodeGen/X86/oddshuffles.ll Sat Oct 6 06:49:31 2018
@@ -1036,7 +1036,7 @@ define void @interleave_24i16_out(<24 x
; SSE42-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
; SSE42-NEXT: movdqa %xmm0, %xmm4
; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3],xmm1[4],xmm4[5,6],xmm1[7]
-; SSE42-NEXT: pshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; SSE42-NEXT: pshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm3[6,7]
; SSE42-NEXT: movdqa %xmm2, %xmm3
; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[0,1,6,7,4,5,6,7,0,1,0,1,6,7,12,13]
@@ -1059,12 +1059,12 @@ define void @interleave_24i16_out(<24 x
; AVX1-NEXT: vmovdqu (%rdi), %ymm1
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0],xmm2[1],xmm1[2,3],xmm2[4],xmm1[5,6],xmm2[7]
-; AVX1-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u]
; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,1,2,1]
; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,5]
; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7]
-; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[2,3,8,9,14,15,4,5,10,11,10,11,8,9,14,15]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[2,3,8,9,14,15,4,5,10,11,u,u,u,u,u,u]
; AVX1-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[0,1,6,7,4,5,6,7,0,1,0,1,6,7,12,13]
; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm5[5,6,7]
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7]
@@ -1202,7 +1202,7 @@ define void @interleave_24i16_in(<24 x i
; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[1,1,2,2]
; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3],xmm3[4,5],xmm4[6],xmm3[7]
; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,2,3,u,u,4,5,6,7,u,u,8,9,10,11]
; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[0,0,0,3]
; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7]
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3
Modified: llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v8.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v8.ll?rev=343919&r1=343918&r2=343919&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v8.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v8.ll Sat Oct 6 06:49:31 2018
@@ -1557,14 +1557,14 @@ define <8 x i16> @shuffle_v8i16_XX4X8acX
;
; SSE41-LABEL: shuffle_v8i16_XX4X8acX:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,4,5,6,7,0,1,4,5,8,9,4,5]
+; SSE41-NEXT: pshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,0,1,4,5,8,9,4,5]
; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; SSE41-NEXT: retq
;
; AVX1-LABEL: shuffle_v8i16_XX4X8acX:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,4,5,6,7,0,1,4,5,8,9,4,5]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,0,1,4,5,8,9,4,5]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
; AVX1-NEXT: retq
Modified: llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll?rev=343919&r1=343918&r2=343919&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll Sat Oct 6 06:49:31 2018
@@ -3732,7 +3732,7 @@ define <16 x i16> @shuffle_v16i16_uu_uu_
; AVX1-LABEL: shuffle_v16i16_uu_uu_04_uu_16_18_20_uu_uu_uu_12_uu_24_26_28_uu:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,4,5,4,5,6,7,0,1,4,5,8,9,4,5]
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,u,u,u,u,u,0,1,4,5,8,9,4,5>
; AVX1-NEXT: vpshufb %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,2,3,3]
@@ -3873,7 +3873,7 @@ define <16 x i16> @shuffle_v16i16_19_20_
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5]
; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,4,5,6,7,8,9,0,1,4,5,10,11]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,u,u,u,0,1,4,5,10,11]
; AVX1-NEXT: vpsrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5,6,7]
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
@@ -3982,7 +3982,7 @@ define <16 x i16> @shuffle_v16i16_03_04_
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5]
; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; AVX1-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,4,5,6,7,8,9,0,1,4,5,10,11]
+; AVX1-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,u,u,0,1,4,5,10,11]
; AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5,6,7]
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
More information about the llvm-commits
mailing list