From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Thu, 4 Jul 2019 10:17:10 +0000 (+0000)
Subject: [X86][AVX1] Combine concat_vectors(pshufd(x,c),pshufd(y,c)) -> vpermilps(concat_vecto... 
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=23d6310164cd55c6cd09b78f67357a1974a85edf;p=llvm

[X86][AVX1] Combine concat_vectors(pshufd(x,c),pshufd(y,c)) -> vpermilps(concat_vectors(x,y),c)

Bitcast v4i32 to v8f32 and back again - it might be worth adding isel patterns for X86PShufd v8i32 on AVX1 targets like we did for X86Blendi to avoid the bitcasts?

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@365125 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 8b3378f0c6c..1b9312d2c1f 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -43551,18 +43551,18 @@ static SDValue combineConcatVectorOps(const SDLoc &DL, MVT VT,
                            DAG.getNode(ISD::CONCAT_VECTORS, DL, VT, Src),
                            Op0.getOperand(1));
       }
-      break;
+      LLVM_FALLTHROUGH;
     case X86ISD::VPERMILPI:
-      // TODO - AVX1 must use VPERMILPI + v8f32 for v8i32 shuffles.
       // TODO - add support for vXf64/vXi64 shuffles.
-      if (!IsSplat && NumOps == 2 && VT == MVT::v8f32 && Subtarget.hasAVX() &&
-          Op0.getOperand(1) == Ops[1].getOperand(1)) {
+      if (!IsSplat && NumOps == 2 && (VT == MVT::v8f32 || VT == MVT::v8i32) &&
+          Subtarget.hasAVX() && Op0.getOperand(1) == Ops[1].getOperand(1)) {
         SmallVector<SDValue, 2> Src;
         for (unsigned i = 0; i != NumOps; ++i)
-          Src.push_back(Ops[i].getOperand(0));
-        return DAG.getNode(Op0.getOpcode(), DL, VT,
-                           DAG.getNode(ISD::CONCAT_VECTORS, DL, VT, Src),
-                           Op0.getOperand(1));
+          Src.push_back(DAG.getBitcast(MVT::v4f32, Ops[i].getOperand(0)));
+        SDValue Res = DAG.getNode(ISD::CONCAT_VECTORS, DL, MVT::v8f32, Src);
+        Res = DAG.getNode(X86ISD::VPERMILPI, DL, MVT::v8f32, Res,
+                          Op0.getOperand(1));
+        return DAG.getBitcast(VT, Res);
       }
       break;
     case X86ISD::PACKUS:
diff --git a/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll b/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
index 6fa470a06ef..aa71d3749ca 100644
--- a/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
+++ b/test/CodeGen/X86/bitcast-int-to-vector-bool-sext.ll
@@ -361,10 +361,9 @@ define <32 x i8> @ext_i32_32i8(i32 %a0) {
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
@@ -646,10 +645,9 @@ define <64 x i8> @ext_i64_64i8(i64 %a0) {
 ; AVX1-NEXT:    vmovq %rdi, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [-1.7939930131212661E-307,-1.7939930131212661E-307,-1.7939930131212661E-307,-1.7939930131212661E-307]
 ; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
@@ -661,10 +659,9 @@ define <64 x i8> @ext_i64_64i8(i64 %a0) {
 ; AVX1-NEXT:    vpxor %xmm5, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,5,5]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,7,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,3,3,6,6,7,7]
 ; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpcmpeqb %xmm4, %xmm2, %xmm2
diff --git a/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll b/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
index e9886a12bd6..00bbaf38366 100644
--- a/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
+++ b/test/CodeGen/X86/bitcast-int-to-vector-bool-zext.ll
@@ -460,10 +460,9 @@ define <32 x i8> @ext_i32_32i8(i32 %a0) {
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
@@ -834,10 +833,9 @@ define <64 x i8> @ext_i64_64i8(i64 %a0) {
 ; AVX1-NEXT:    vmovq %rdi, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm1[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm1[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [-1.7939930131212661E-307,-1.7939930131212661E-307,-1.7939930131212661E-307,-1.7939930131212661E-307]
 ; AVX1-NEXT:    vandps %ymm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
@@ -854,10 +852,9 @@ define <64 x i8> @ext_i64_64i8(i64 %a0) {
 ; AVX1-NEXT:    vpand %xmm6, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,5,5]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,7,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[2,2,3,3,6,6,7,7]
 ; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpcmpeqb %xmm4, %xmm2, %xmm2
diff --git a/test/CodeGen/X86/bitcast-int-to-vector-bool.ll b/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
index 5a2b4b2494e..c2f7938a8df 100644
--- a/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
+++ b/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
@@ -200,10 +200,9 @@ define <32 x i1> @bitcast_i32_32i1(i32 %a0) {
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
diff --git a/test/CodeGen/X86/vector-sext-widen.ll b/test/CodeGen/X86/vector-sext-widen.ll
index 2dc4817b56e..08e8b514c60 100644
--- a/test/CodeGen/X86/vector-sext-widen.ll
+++ b/test/CodeGen/X86/vector-sext-widen.ll
@@ -2554,10 +2554,9 @@ define <32 x i8> @load_sext_32i1_to_32i8(<32 x i1> *%ptr) nounwind readnone {
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
diff --git a/test/CodeGen/X86/vector-sext.ll b/test/CodeGen/X86/vector-sext.ll
index 2d51df117fa..dd96924be6b 100644
--- a/test/CodeGen/X86/vector-sext.ll
+++ b/test/CodeGen/X86/vector-sext.ll
@@ -2554,10 +2554,9 @@ define <32 x i8> @load_sext_32i1_to_32i8(<32 x i1> *%ptr) nounwind readnone {
 ; AVX1-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
diff --git a/test/CodeGen/X86/vector-shuffle-256-v16.ll b/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 843e06e88cd..0be1fec5ecd 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -496,11 +496,10 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,2,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
@@ -532,11 +531,10 @@ define <16 x i16> @shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_1
 ; AVX1-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,7,7,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,2,2,2,6,6,6,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
@@ -1123,13 +1121,12 @@ define <16 x i16> @shuffle_v16i16_00_16_00_16_00_16_00_16_00_16_00_16_00_16_00_1
 define <16 x i16> @shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
@@ -1246,11 +1243,10 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_08:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,1]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,1,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,1,4,4,4,5]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_08:
@@ -1282,11 +1278,10 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_08:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,0,2,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,2,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_08:
@@ -1318,11 +1313,10 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_08:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[0,0,3,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,1,0,4,4,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_08:
@@ -2233,10 +2227,9 @@ define <16 x i16> @shuffle_v16i16_06_07_04_05_02_03_00_09_14_15_12_13_10_11_08_0
 ; AVX1-LABEL: shuffle_v16i16_06_07_04_05_02_03_00_09_14_15_12_13_10_11_08_09:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6,7]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_06_07_04_05_02_03_00_09_14_15_12_13_10_11_08_09:
@@ -2421,10 +2414,9 @@ define <16 x i16> @shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
@@ -2483,10 +2475,9 @@ define <16 x i16> @shuffle_v16i16_02_03_00_01_06_07_04_13_10_11_08_09_14_15_12_1
 ; AVX1-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_13_10_11_08_09_14_15_12_13:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,3,2]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,0,3,2]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_13_10_11_08_09_14_15_12_13:
@@ -2511,10 +2502,9 @@ define <16 x i16> @shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5],xmm0[6,7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,3,0,2,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[2,3,0,2,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,1,3,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
@@ -4382,13 +4372,12 @@ define <16 x i16> @shuffle_v16i16_9_9_9_9_9_9_9_9_u_u_u_u_u_u_u_u(<16 x i16> %a,
 define <16 x i16> @shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25(<16 x i16> %a0, <16 x i16> %a1) {
 ; AVX1-LABEL: shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_02_18_03_19_00_16_01_17_10_26_11_27_08_24_09_25:
@@ -4448,11 +4437,10 @@ define <16 x i16> @shuffle_v16i16_04_06_07_uu_uu_06_07_05_12_14_15_uu_uu_14_15_1
 ; AVX1-LABEL: shuffle_v16i16_04_06_07_uu_uu_06_07_05_12_14_15_uu_uu_14_15_13:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,6,7,5]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,7,5]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_04_06_07_uu_uu_06_07_05_12_14_15_uu_uu_14_15_13:
diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll
index 8dfcffbe0a4..8dd211b5566 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -1680,13 +1680,12 @@ define <32 x i8> @shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_16_48_16_48_16_48_16_48_16_48_16_48_16_48_16_48:
diff --git a/test/CodeGen/X86/vector-shuffle-256-v4.ll b/test/CodeGen/X86/vector-shuffle-256-v4.ll
index 3717d05e9e5..3603407f2b1 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -1453,26 +1453,12 @@ define <4 x i64> @shuffle_v4i64_1z3z(<4 x i64> %a, <4 x i64> %b) {
 }
 
 define <4 x i64> @shuffle_v4i64_1032_v2i64(<2 x i64> %a, <2 x i64> %b) {
-; AVX1-LABEL: shuffle_v4i64_1032_v2i64:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: shuffle_v4i64_1032_v2i64:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; AVX2-NEXT:    retq
-;
-; AVX512VL-LABEL: shuffle_v4i64_1032_v2i64:
-; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX512VL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
-; AVX512VL-NEXT:    retq
+; ALL-LABEL: shuffle_v4i64_1032_v2i64:
+; ALL:       # %bb.0:
+; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
+; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
+; ALL-NEXT:    retq
   %1 = shufflevector <2 x i64> %a, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
   %2 = shufflevector <2 x i64> %b, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
   %3 = shufflevector <2 x i64> %1, <2 x i64> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
diff --git a/test/CodeGen/X86/vector-shuffle-256-v8.ll b/test/CodeGen/X86/vector-shuffle-256-v8.ll
index 9969167a10d..795ca11ca26 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -2557,19 +2557,12 @@ define <8 x i32> @shuffle_v8i32_uuuuuu7u(<8 x i32> %a, <8 x i32> %b) nounwind {
 }
 
 define <8 x i32> @shuffle_v8i32_32107654_v4i32(<4 x i32> %a, <4 x i32> %b) {
-; AVX1-LABEL: shuffle_v8i32_32107654_v4i32:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,2,1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2OR512VL-LABEL: shuffle_v8i32_32107654_v4i32:
-; AVX2OR512VL:       # %bb.0:
-; AVX2OR512VL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; AVX2OR512VL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2OR512VL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
-; AVX2OR512VL-NEXT:    retq
+; ALL-LABEL: shuffle_v8i32_32107654_v4i32:
+; ALL:       # %bb.0:
+; ALL-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
+; ALL-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; ALL-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
+; ALL-NEXT:    retq
   %1 = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %2 = shufflevector <4 x i32> %b, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %3 = shufflevector <4 x i32> %1, <4 x i32> %2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>