[x86] Delete still more piles of complex code now that we have a good

systematic lowering of v8i16.

This required a slight strategy shift to prefer unpack lowerings in more
places. While this isn't a cut-and-dry win in every case, it is in the
overwhelming majority. There are only a few places where the old
lowering would probably be a touch faster, and then only by a small
margin.

In some cases, this is yet another significant improvement.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229859 91177308-0d34-0410-b5e6-96231b3b80d8

6 files changed, 121 insertions, 245 deletions

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 45128fe032a..596d72704b6 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -9518,44 +9518,6 @@ static SDValue lowerV8I16SingleInputVectorShuffle(
   return V;
 }
 
-/// \brief Detect whether the mask pattern should be lowered through
-/// interleaving.
-///
-/// This essentially tests whether viewing the mask as an interleaving of two
-/// sub-sequences reduces the cross-input traffic of a blend operation. If so,
-/// lowering it through interleaving is a significantly better strategy.
-static bool shouldLowerAsInterleaving(ArrayRef<int> Mask) {
-  int NumEvenInputs[2] = {0, 0};
-  int NumOddInputs[2] = {0, 0};
-  int NumLoInputs[2] = {0, 0};
-  int NumHiInputs[2] = {0, 0};
-  for (int i = 0, Size = Mask.size(); i < Size; ++i) {
-    if (Mask[i] < 0)
-      continue;
-
-    int InputIdx = Mask[i] >= Size;
-
-    if (i < Size / 2)
-      ++NumLoInputs[InputIdx];
-    else
-      ++NumHiInputs[InputIdx];
-
-    if ((i % 2) == 0)
-      ++NumEvenInputs[InputIdx];
-    else
-      ++NumOddInputs[InputIdx];
-  }
-
-  // The minimum number of cross-input results for both the interleaved and
-  // split cases. If interleaving results in fewer cross-input results, return
-  // true.
-  int InterleavedCrosses = std::min(NumEvenInputs[1] + NumOddInputs[0],
-                                    NumEvenInputs[0] + NumOddInputs[1]);
-  int SplitCrosses = std::min(NumLoInputs[1] + NumHiInputs[0],
-                              NumLoInputs[0] + NumHiInputs[1]);
-  return InterleavedCrosses < SplitCrosses;
-}
-
 /// \brief Helper to form a PSHUFB-based shuffle+blend.
 static SDValue lowerVectorShuffleAsPSHUFB(SDLoc DL, MVT VT, SDValue V1,
                                           SDValue V2, ArrayRef<int> Mask,
@@ -9691,43 +9653,16 @@ static SDValue lowerV8I16VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
           lowerVectorShuffleAsBitBlend(DL, MVT::v8i16, V1, V2, Mask, DAG))
     return BitBlend;
 
-  // Check whether an interleaving lowering is likely to be more efficient.
-  // This isn't perfect but it is a strong heuristic that tends to work well on
-  // the kinds of shuffles that show up in practice.
-  //
-  // FIXME: Handle 1x, 2x, and 4x interleaving.
-  if (shouldLowerAsInterleaving(Mask)) {
-    // FIXME: Figure out whether we should pack these into the low or high
-    // halves.
-
-    int EMask[8], OMask[8];
-    for (int i = 0; i < 4; ++i) {
-      EMask[i] = Mask[2*i];
-      OMask[i] = Mask[2*i + 1];
-      EMask[i + 4] = -1;
-      OMask[i + 4] = -1;
-    }
-
-    SDValue Evens = DAG.getVectorShuffle(MVT::v8i16, DL, V1, V2, EMask);
-    SDValue Odds = DAG.getVectorShuffle(MVT::v8i16, DL, V1, V2, OMask);
-
-    return DAG.getNode(X86ISD::UNPCKL, DL, MVT::v8i16, Evens, Odds);
-  }
-
-  // Try to lower by permuting the inputs into an unpack instruction unless we
-  // have direct support for blending.
-  if (!IsBlendSupported) {
-    if (SDValue Unpack =
-            lowerVectorShuffleAsUnpack(MVT::v8i16, DL, V1, V2, Mask, DAG))
-      return Unpack;
+  if (SDValue Unpack =
+          lowerVectorShuffleAsUnpack(MVT::v8i16, DL, V1, V2, Mask, DAG))
+    return Unpack;
 
-    // If we can use PSHUFB, that will be better as it can both shuffle and set
-    // up an efficient blend.
-    if (Subtarget->hasSSSE3()) {
-      bool V1InUse, V2InUse;
-      return lowerVectorShuffleAsPSHUFB(DL, MVT::v8i16, V1, V2, Mask, DAG,
-                                        V1InUse, V2InUse);
-    }
+  // If we can't directly blend but can use PSHUFB, that will be better as it
+  // can both shuffle and set up the inefficient blend.
+  if (!IsBlendSupported && Subtarget->hasSSSE3()) {
+    bool V1InUse, V2InUse;
+    return lowerVectorShuffleAsPSHUFB(DL, MVT::v8i16, V1, V2, Mask, DAG,
+                                      V1InUse, V2InUse);
   }
 
   // We can always bit-blend if we have to so the fallback strategy is to
diff --git a/test/CodeGen/X86/vector-shuffle-128-v16.ll b/test/CodeGen/X86/vector-shuffle-128-v16.ll
index 3cc52fb8c6f..8f1f2fb7833 100644
--- a/test/CodeGen/X86/vector-shuffle-128-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-128-v16.ll
@@ -1094,33 +1094,33 @@ define <16 x i8> @shuffle_v16i8_uu_10_02_07_22_14_07_02_18_03_01_14_18_09_11_00(
 ; SSE2:       # BB#0: # %entry
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
 ; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[3,1,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm4[1,2,2,3,4,5,6,7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,1,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,1,4,5,6,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,0,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,0,0,65535]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,4]
-; SSE2-NEXT:    pand %xmm2, %xmm3
-; SSE2-NEXT:    pandn %xmm0, %xmm2
-; SSE2-NEXT:    por %xmm3, %xmm2
-; SSE2-NEXT:    packuswb %xmm2, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,2,2,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,7,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,65535,0,0,65535]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,1,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,4]
+; SSE2-NEXT:    pand %xmm5, %xmm2
+; SSE2-NEXT:    pandn %xmm4, %xmm5
+; SSE2-NEXT:    por %xmm2, %xmm5
+; SSE2-NEXT:    psrlq $16, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,1,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,4]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE2-NEXT:    packuswb %xmm5, %xmm2
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,255,255,255,0,255,255,255,0,255,255,255]
-; SSE2-NEXT:    pand %xmm0, %xmm4
+; SSE2-NEXT:    pand %xmm0, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,7]
 ; SSE2-NEXT:    pandn %xmm1, %xmm0
-; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v16i8_uu_10_02_07_22_14_07_02_18_03_01_14_18_09_11_00:
diff --git a/test/CodeGen/X86/vector-shuffle-128-v8.ll b/test/CodeGen/X86/vector-shuffle-128-v8.ll
index 84b47faafaa..f6bf7884a63 100644
--- a/test/CodeGen/X86/vector-shuffle-128-v8.ll
+++ b/test/CodeGen/X86/vector-shuffle-128-v8.ll
@@ -949,40 +949,19 @@ define <8 x i16> @shuffle_v8i16_0c1d6879(<8 x i16> %a, <8 x i16> %b) {
 }
 
 define <8 x i16> @shuffle_v8i16_109832ba(<8 x i16> %a, <8 x i16> %b) {
-; SSE2-LABEL: shuffle_v8i16_109832ba:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: shuffle_v8i16_109832ba:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
-; SSSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: shuffle_v8i16_109832ba:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    pshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,2,3,0,1,8,9,10,11,6,7,4,5]
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[2,3,0,1,4,5,6,7,6,7,4,5,4,5,6,7]
-; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
-; SSE41-NEXT:    retq
-;
-; AVX1-LABEL: shuffle_v8i16_109832ba:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,2,3,0,1,8,9,10,11,6,7,4,5]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,0,1,4,5,6,7,6,7,4,5,4,5,6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
-; AVX1-NEXT:    retq
+; SSE-LABEL: shuffle_v8i16_109832ba:
+; SSE:       # BB#0:
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
+; SSE-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8i16_109832ba:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,2,3,0,1,8,9,10,11,6,7,4,5]
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3,0,1,4,5,6,7,6,7,4,5,4,5,6,7]
-; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX2-NEXT:    retq
+; AVX-LABEL: shuffle_v8i16_109832ba:
+; AVX:       # BB#0:
+; AVX-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[2,0,3,1,4,5,6,7]
+; AVX-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,7,5]
+; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 1, i32 0, i32 9, i32 8, i32 3, i32 2, i32 11, i32 10>
   ret <8 x i16> %shuffle
 }
@@ -1017,42 +996,21 @@ define <8 x i16> @shuffle_v8i16_c4d5e6f7(<8 x i16> %a, <8 x i16> %b) {
 }
 
 define <8 x i16> @shuffle_v8i16_0213cedf(<8 x i16> %a, <8 x i16> %b) {
-; SSE2-LABEL: shuffle_v8i16_0213cedf:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: shuffle_v8i16_0213cedf:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
-; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: shuffle_v8i16_0213cedf:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
-; SSE41-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
-; SSE41-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
-; SSE41-NEXT:    retq
-;
-; AVX1-LABEL: shuffle_v8i16_0213cedf:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
-; AVX1-NEXT:    retq
+; SSE-LABEL: shuffle_v8i16_0213cedf:
+; SSE:       # BB#0:
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    retq
 ;
-; AVX2-LABEL: shuffle_v8i16_0213cedf:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
-; AVX2-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
-; AVX2-NEXT:    retq
+; AVX-LABEL: shuffle_v8i16_0213cedf:
+; AVX:       # BB#0:
+; AVX-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
+; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
+; AVX-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 0, i32 2, i32 1, i32 3, i32 12, i32 14, i32 13, i32 15>
   ret <8 x i16> %shuffle
 }
@@ -2161,32 +2119,19 @@ define <8 x i16> @shuffle_v8i16_0123456z(<8 x i16> %a) {
 }
 
 define <8 x i16> @shuffle_v8i16_fu3ucc5u(<8 x i16> %a, <8 x i16> %b) {
-; SSE2-LABEL: shuffle_v8i16_fu3ucc5u:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,4,4]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: shuffle_v8i16_fu3ucc5u:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,4,4]
-; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm0
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: shuffle_v8i16_fu3ucc5u:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5,6],xmm1[7]
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[14,15,14,15,6,7,6,7,8,9,8,9,10,11,14,15]
-; SSE41-NEXT:    retq
+; SSE-LABEL: shuffle_v8i16_fu3ucc5u:
+; SSE:       # BB#0:
+; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,4,4]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v8i16_fu3ucc5u:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5,6],xmm1[7]
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[14,15,14,15,6,7,6,7,8,9,8,9,10,11,14,15]
+; AVX-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; AVX-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,4,4]
+; AVX-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 15, i32 undef, i32 3, i32 undef, i32 12, i32 12, i32 5, i32 undef>
   ret <8 x i16> %shuffle
diff --git a/test/CodeGen/X86/vector-shuffle-256-v16.ll b/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 1cb0354285e..9db201548cc 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -150,12 +150,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,4,4,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5,6],xmm1[7]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,0,1,2,3]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -175,8 +172,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_00_09_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2,3,4,5,6,7]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,2,3,0,1]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,0,1,6,7,0,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
@@ -196,8 +193,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_00_10_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,0,1,10,11,0,1,0,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
@@ -216,8 +213,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_0
 ; AVX1-LABEL: shuffle_v16i16_00_00_00_00_11_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
-; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1]
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,0,1,0,1,0,1,14,15,0,1,0,1,0,1]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
@@ -690,9 +687,8 @@ define <16 x i16> @shuffle_v16i16_00_01_18_19_20_21_06_07_08_09_26_27_12_13_30_3
 define <16 x i16> @shuffle_v16i16_00_16_00_16_00_16_00_16_00_16_00_16_00_16_00_16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_00_16_00_16_00_16_00_16:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; 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, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -709,15 +705,13 @@ 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:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; 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:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
@@ -734,11 +728,13 @@ define <16 x i16> @shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_1
 ; AVX1-LABEL: shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_15:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,3,0,1]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -754,15 +750,17 @@ define <16 x i16> @shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_1
 define <16 x i16> @shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm3 = xmm3[3,2,1,0,4,5,6,7]
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
@@ -777,16 +775,13 @@ define <16 x i16> @shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_1
 define <16 x i16> @shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,1]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
-; AVX1-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [12,13,8,9,4,5,0,1,14,15,10,11,6,7,2,3]
+; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
@@ -1434,12 +1429,11 @@ define <16 x i16> @shuffle_v16i16_02_03_zz_zz_06_07_zz_zz_10_11_zz_zz_14_15_zz_z
 define <16 x i16> @shuffle_v16i16_16_zz_zz_zz_17_zz_zz_zz_18_zz_zz_zz_19_zz_zz_zz(<16 x i16> %a) {
 ; AVX1-LABEL: shuffle_v16i16_16_zz_zz_zz_17_zz_zz_zz_18_zz_zz_zz_19_zz_zz_zz:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm0[4,5,2,3,4,5,6,7,6,7,10,11,4,5,6,7]
 ; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1,2,3],xmm1[4],xmm2[5,6,7]
+; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v16i16_16_zz_zz_zz_17_zz_zz_zz_18_zz_zz_zz_19_zz_zz_zz:
@@ -1453,9 +1447,8 @@ define <16 x i16> @shuffle_v16i16_16_zz_zz_zz_17_zz_zz_zz_18_zz_zz_zz_19_zz_zz_z
 define <16 x i16> @shuffle_v16i16_16_zz_17_zz_18_zz_19_zz_20_zz_21_zz_22_zz_22_zz(<16 x i16> %a) {
 ; AVX1-LABEL: shuffle_v16i16_16_zz_17_zz_18_zz_19_zz_20_zz_21_zz_22_zz_22_zz:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm1 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
diff --git a/test/CodeGen/X86/vector-trunc.ll b/test/CodeGen/X86/vector-trunc.ll
index 80f3dee4332..a3360150338 100644
--- a/test/CodeGen/X86/vector-trunc.ll
+++ b/test/CodeGen/X86/vector-trunc.ll
@@ -82,16 +82,18 @@ define <8 x i16> @trunc2x4i32(<4 x i32> %a, <4 x i32> %b) {
 ;
 ; SSE41-LABEL: trunc2x4i32:
 ; SSE41:       # BB#0: # %entry
-; SSE41-NEXT:    pshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,4,5,6,7,0,1,4,5,8,9,12,13]
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
+; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE41-NEXT:    pshufb %xmm2, %xmm1
+; SSE41-NEXT:    pshufb %xmm2, %xmm0
+; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: trunc2x4i32:
 ; AVX:       # BB#0: # %entry
-; AVX-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,4,5,6,7,0,1,4,5,8,9,12,13]
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
+; AVX-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX-NEXT:    retq
 
 
diff --git a/test/CodeGen/X86/widen_shuffle-1.ll b/test/CodeGen/X86/widen_shuffle-1.ll
index 546ce4502f0..2aa870f16eb 100644
--- a/test/CodeGen/X86/widen_shuffle-1.ll
+++ b/test/CodeGen/X86/widen_shuffle-1.ll
@@ -68,9 +68,10 @@ entry:
 define <8 x i8> @shuf4(<4 x i8> %a, <4 x i8> %b) nounwind readnone {
 ; CHECK-LABEL: shuf4:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    pshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,4,5,6,7,0,1,4,5,8,9,12,13]
-; CHECK-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; CHECK-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; CHECK-NEXT:    pshufb %xmm2, %xmm1
+; CHECK-NEXT:    pshufb %xmm2, %xmm0
+; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; CHECK-NEXT:    retl
   %vshuf = shufflevector <4 x i8> %a, <4 x i8> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   ret <8 x i8> %vshuf