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

llvm-svn: 358170

2 files changed, 21 insertions, 83 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index a896f24c365..6319c9cc223 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -33225,7 +33225,8 @@ bool X86TargetLowering::SimplifyDemandedVectorEltsForTargetNode(
     }
     break;
   }
-  case X86ISD::PSHUFB: {
+  case X86ISD::PSHUFB:
+  case X86ISD::VPERMILPV: {
     // TODO - simplify other variable shuffle masks.
     SDValue Mask = Op.getOperand(1);
     APInt MaskUndef, MaskZero;
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
index 1afbc2adf75..d87729b123e 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
@@ -188,41 +188,10 @@ define <8 x float> @combine_vpermilvar_8f32_movshdup(<8 x float> %a0) {
   ret <8 x float> %1
 }
 define <8 x float> @demandedelts_vpermilvar_8f32_movshdup(<8 x float> %a0, i32 %a1) {
-; X86-LABEL: demandedelts_vpermilvar_8f32_movshdup:
-; X86:       # %bb.0:
-; X86-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %ymm1
-; X86-NEXT:    vblendps {{.*#+}} ymm1 = mem[0,1,2,3,4,5,6],ymm1[7]
-; X86-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X86-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,6]
-; X86-NEXT:    retl
-;
-; X64-AVX1-LABEL: demandedelts_vpermilvar_8f32_movshdup:
-; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vmovd %edi, %xmm1
-; X64-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; X64-AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; X64-AVX1-NEXT:    vblendps {{.*#+}} ymm1 = mem[0,1,2,3,4,5,6],ymm1[7]
-; X64-AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,6]
-; X64-AVX1-NEXT:    retq
-;
-; X64-AVX2-LABEL: demandedelts_vpermilvar_8f32_movshdup:
-; X64-AVX2:       # %bb.0:
-; X64-AVX2-NEXT:    vmovd %edi, %xmm1
-; X64-AVX2-NEXT:    vpbroadcastd %xmm1, %ymm1
-; X64-AVX2-NEXT:    vpblendd {{.*#+}} ymm1 = mem[0,1,2,3,4,5,6],ymm1[7]
-; X64-AVX2-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,6]
-; X64-AVX2-NEXT:    retq
-;
-; X64-AVX512-LABEL: demandedelts_vpermilvar_8f32_movshdup:
-; X64-AVX512:       # %bb.0:
-; X64-AVX512-NEXT:    vmovd %edi, %xmm1
-; X64-AVX512-NEXT:    vpbroadcastd %xmm1, %ymm1
-; X64-AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = mem[0,1,2,3,4,5,6],ymm1[7]
-; X64-AVX512-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X64-AVX512-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6,6]
-; X64-AVX512-NEXT:    retq
+; CHECK-LABEL: demandedelts_vpermilvar_8f32_movshdup:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
+; CHECK-NEXT:    ret{{[l|q]}}
   %1 = insertelement <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 undef, i32 5, i32 7, i32 7>, i32 %a1, i32 7
   %2 = tail call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> %1)
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 6>
@@ -238,55 +207,23 @@ define <8 x float> @combine_vpermilvar_8f32_movsldup(<8 x float> %a0) {
   ret <8 x float> %1
 }
 define <8 x float> @demandedelts_vpermilvar_8f32_movsldup(<8 x float> %a0, i32 %a1) {
-; X86-AVX1-LABEL: demandedelts_vpermilvar_8f32_movsldup:
-; X86-AVX1:       # %bb.0:
-; X86-AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <u,0,2,2,4,4,6,6>
-; X86-AVX1-NEXT:    vpinsrd $0, {{[0-9]+}}(%esp), %xmm1, %xmm1
-; X86-AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
-; X86-AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X86-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
-; X86-AVX1-NEXT:    retl
-;
-; X86-AVX2-LABEL: demandedelts_vpermilvar_8f32_movsldup:
-; X86-AVX2:       # %bb.0:
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],mem[1,2,3,4,5,6,7]
-; X86-AVX2-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X86-AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
-; X86-AVX2-NEXT:    retl
-;
-; X86-AVX512-LABEL: demandedelts_vpermilvar_8f32_movsldup:
-; X86-AVX512:       # %bb.0:
-; X86-AVX512-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X86-AVX512-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0],mem[1,2,3,4,5,6,7]
-; X86-AVX512-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X86-AVX512-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
-; X86-AVX512-NEXT:    retl
-;
-; X64-AVX1-LABEL: demandedelts_vpermilvar_8f32_movsldup:
-; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <u,0,2,2,4,4,6,6>
-; X64-AVX1-NEXT:    vpinsrd $0, %edi, %xmm1, %xmm1
-; X64-AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
-; X64-AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X64-AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
-; X64-AVX1-NEXT:    retq
+; AVX1-LABEL: demandedelts_vpermilvar_8f32_movsldup:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovaps {{.*#+}} xmm1 = <u,0,2,2,4,4,6,6>
+; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
+; AVX1-NEXT:    ret{{[l|q]}}
 ;
-; X64-AVX2-LABEL: demandedelts_vpermilvar_8f32_movsldup:
-; X64-AVX2:       # %bb.0:
-; X64-AVX2-NEXT:    vmovd %edi, %xmm1
-; X64-AVX2-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2,3,4,5,6,7]
-; X64-AVX2-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X64-AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
-; X64-AVX2-NEXT:    retq
+; AVX2-LABEL: demandedelts_vpermilvar_8f32_movsldup:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
+; AVX2-NEXT:    ret{{[l|q]}}
 ;
-; X64-AVX512-LABEL: demandedelts_vpermilvar_8f32_movsldup:
-; X64-AVX512:       # %bb.0:
-; X64-AVX512-NEXT:    vmovd %edi, %xmm1
-; X64-AVX512-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2,3,4,5,6,7]
-; X64-AVX512-NEXT:    vpermilps %ymm1, %ymm0, %ymm0
-; X64-AVX512-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,1,2,3,4,5,6,7]
-; X64-AVX512-NEXT:    retq
+; AVX512-LABEL: demandedelts_vpermilvar_8f32_movsldup:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6]
+; AVX512-NEXT:    ret{{[l|q]}}
   %1 = insertelement <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>, i32 %a1, i32 0
   %2 = tail call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> %1)
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 1, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>