[x86] Make the v8i16 clever single-input shuffle lowering usable for

repeated 128-bit lane shuffles of wider vector types and use it to lower
256-bit v16i16 vector shuffles where applicable.

This should let us perfectly lowering the pattern of pshuflw and pshufhw
even for AVX2 256-bit patterns.

I've not added AVX-512 support, but it should be trivial for someone
working on that to wire up.

Note that currently this generates bad, long shuffle chains because we
don't combine 256-bit target shuffles. The subsequent patches will fix
that.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230751 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 1bc68f3eaf771c2944dbbc7083c894f0d4e27c92..bafb661ff0528159815cc141ad4bc9a37be3acc6 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -7689,10 +7689,18 @@ static SDValue lowerV4I32VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
 /// The exact breakdown of how to form these dword pairs and align them on the
 /// correct sides is really tricky. See the comments within the function for
 /// more of the details.
+///
+/// This code also handles repeated 128-bit lanes of v8i16 shuffles, but each
+/// lane must shuffle the *exact* same way. In fact, you must pass a v8 Mask to
+/// this routine for it to work correctly. To shuffle a 256-bit or 512-bit i16
+/// vector, form the analogous 128-bit 8-element Mask.
 static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
-    SDLoc DL, SDValue V, MutableArrayRef<int> Mask,
+    SDLoc DL, MVT VT, SDValue V, MutableArrayRef<int> Mask,
     const X86Subtarget *Subtarget, SelectionDAG &DAG) {
-  assert(V.getSimpleValueType() == MVT::v8i16 && "Bad input type!");
+  assert(VT.getScalarType() == MVT::i16 && "Bad input type!");
+  MVT PSHUFDVT = MVT::getVectorVT(MVT::i32, VT.getVectorNumElements() / 2);
+
+  assert(Mask.size() == 8 && "Shuffle mask length doen't match!");
   MutableArrayRef<int> LoMask = Mask.slice(0, 4);
   MutableArrayRef<int> HiMask = Mask.slice(4, 4);
 
@@ -7845,9 +7853,9 @@ static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
     int PSHUFDMask[] = {0, 1, 2, 3};
     PSHUFDMask[ADWord] = BDWord;
     PSHUFDMask[BDWord] = ADWord;
-    V = DAG.getNode(ISD::BITCAST, DL, MVT::v8i16,
-                    DAG.getNode(X86ISD::PSHUFD, DL, MVT::v4i32,
-                                DAG.getNode(ISD::BITCAST, DL, MVT::v4i32, V),
+    V = DAG.getNode(ISD::BITCAST, DL, VT,
+                    DAG.getNode(X86ISD::PSHUFD, DL, PSHUFDVT,
+                                DAG.getNode(ISD::BITCAST, DL, PSHUFDVT, V),
                                 getV4X86ShuffleImm8ForMask(PSHUFDMask, DAG)));
 
     // Adjust the mask to match the new locations of A and B.
@@ -7859,7 +7867,7 @@ static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
 
     // Recurse back into this routine to re-compute state now that this isn't
     // a 3 and 1 problem.
-    return lowerV8I16GeneralSingleInputVectorShuffle(DL, V, Mask, Subtarget,
+    return lowerV8I16GeneralSingleInputVectorShuffle(DL, VT, V, Mask, Subtarget,
                                                      DAG);
   };
   if ((NumLToL == 3 && NumHToL == 1) || (NumLToL == 1 && NumHToL == 3))
@@ -8083,15 +8091,15 @@ static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
   // Now enact all the shuffles we've computed to move the inputs into their
   // target half.
   if (!isNoopShuffleMask(PSHUFLMask))
-    V = DAG.getNode(X86ISD::PSHUFLW, DL, MVT::v8i16, V,
+    V = DAG.getNode(X86ISD::PSHUFLW, DL, VT, V,
                     getV4X86ShuffleImm8ForMask(PSHUFLMask, DAG));
   if (!isNoopShuffleMask(PSHUFHMask))
-    V = DAG.getNode(X86ISD::PSHUFHW, DL, MVT::v8i16, V,
+    V = DAG.getNode(X86ISD::PSHUFHW, DL, VT, V,
                     getV4X86ShuffleImm8ForMask(PSHUFHMask, DAG));
   if (!isNoopShuffleMask(PSHUFDMask))
-    V = DAG.getNode(ISD::BITCAST, DL, MVT::v8i16,
-                    DAG.getNode(X86ISD::PSHUFD, DL, MVT::v4i32,
-                                DAG.getNode(ISD::BITCAST, DL, MVT::v4i32, V),
+    V = DAG.getNode(ISD::BITCAST, DL, VT,
+                    DAG.getNode(X86ISD::PSHUFD, DL, PSHUFDVT,
+                                DAG.getNode(ISD::BITCAST, DL, PSHUFDVT, V),
                                 getV4X86ShuffleImm8ForMask(PSHUFDMask, DAG)));
 
   // At this point, each half should contain all its inputs, and we can then
@@ -8105,7 +8113,7 @@ static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
 
   // Do a half shuffle for the low mask.
   if (!isNoopShuffleMask(LoMask))
-    V = DAG.getNode(X86ISD::PSHUFLW, DL, MVT::v8i16, V,
+    V = DAG.getNode(X86ISD::PSHUFLW, DL, VT, V,
                     getV4X86ShuffleImm8ForMask(LoMask, DAG));
 
   // Do a half shuffle with the high mask after shifting its values down.
@@ -8113,7 +8121,7 @@ static SDValue lowerV8I16GeneralSingleInputVectorShuffle(
     if (M >= 0)
       M -= 4;
   if (!isNoopShuffleMask(HiMask))
-    V = DAG.getNode(X86ISD::PSHUFHW, DL, MVT::v8i16, V,
+    V = DAG.getNode(X86ISD::PSHUFHW, DL, VT, V,
                     getV4X86ShuffleImm8ForMask(HiMask, DAG));
 
   return V;
@@ -8232,8 +8240,8 @@ static SDValue lowerV8I16VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
                                                         Mask, Subtarget, DAG))
       return Rotate;
 
-    return lowerV8I16GeneralSingleInputVectorShuffle(DL, V1, Mask, Subtarget,
-                                                     DAG);
+    return lowerV8I16GeneralSingleInputVectorShuffle(DL, MVT::v8i16, V1, Mask,
+                                                     Subtarget, DAG);
   }
 
   assert(std::any_of(Mask.begin(), Mask.end(), isV1) &&
@@ -9557,6 +9565,15 @@ static SDValue lowerV16I16VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
       return lowerVectorShuffleAsLanePermuteAndBlend(DL, MVT::v16i16, V1, V2,
                                                      Mask, DAG);
 
+    SmallVector<int, 8> RepeatedMask;
+    if (is128BitLaneRepeatedShuffleMask(MVT::v16i16, Mask, RepeatedMask)) {
+      // As this is a single-input shuffle, the repeated mask should be
+      // a strictly valid v8i16 mask that we can pass through to the v8i16
+      // lowering to handle even the v16 case.
+      return lowerV8I16GeneralSingleInputVectorShuffle(
+          DL, MVT::v16i16, V1, RepeatedMask, Subtarget, DAG);
+    }
+
     SDValue PSHUFBMask[32];
     for (int i = 0; i < 16; ++i) {
       if (Mask[i] == -1) {

diff --git a/test/CodeGen/X86/vector-shuffle-256-v16.ll b/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 13e3044260d06bc03161af788d809d11d7edaf4c..f7f8044b06f300974adf8fbfb84d0b40273d05ba 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -159,7 +159,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_0
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,3,4,5,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
 ; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
@@ -321,7 +323,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -339,7 +343,9 @@ define <16 x i16> @shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_1
 ;
 ; AVX2-LABEL: shuffle_v16i16_07_07_07_07_07_07_07_07_15_15_15_15_15_15_15_15:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,14,15,14,15,14,15,14,15,14,15,14,15,14,15,30,31,30,31,30,31,30,31,30,31,30,31,30,31,30,31]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,3,2,3,4,7,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
   ret <16 x i16> %shuffle
@@ -358,7 +364,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_1
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_08_08_08_08_12_12_12_12:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12>
   ret <16 x i16> %shuffle
@@ -377,7 +384,8 @@ define <16 x i16> @shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_1
 ;
 ; AVX2-LABEL: shuffle_v16i16_03_03_03_03_07_07_07_07_11_11_11_11_15_15_15_15:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,6,7,6,7,6,7,14,15,14,15,14,15,14,15,22,23,22,23,22,23,22,23,30,31,30,31,30,31,30,31]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,3,3,3,4,5,6,7,11,11,11,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,7,7,7,8,9,10,11,15,15,15,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 7, i32 7, i32 7, i32 7, i32 11, i32 11, i32 11, i32 11, i32 15, i32 15, i32 15, i32 15>
   ret <16 x i16> %shuffle
@@ -396,7 +404,8 @@ define <16 x i16> @shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_1
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_02_02_04_04_06_06_08_08_10_10_12_12_14_14:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5,8,9,8,9,12,13,12,13,16,17,16,17,20,21,20,21,24,25,24,25,28,29,28,29]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,2,2,4,5,6,7,8,8,10,10,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,6,6,8,9,10,11,12,12,14,14]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6, i32 8, i32 8, i32 10, i32 10, i32 12, i32 12, i32 14, i32 14>
   ret <16 x i16> %shuffle
@@ -415,7 +424,8 @@ define <16 x i16> @shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_1
 ;
 ; AVX2-LABEL: shuffle_v16i16_01_01_03_03_05_05_07_07_09_09_11_11_13_13_15_15:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7,10,11,10,11,14,15,14,15,18,19,18,19,22,23,22,23,26,27,26,27,30,31,30,31]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7,9,9,11,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,5,5,7,7,8,9,10,11,13,13,15,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7, i32 9, i32 9, i32 11, i32 11, i32 13, i32 13, i32 15, i32 15>
   ret <16 x i16> %shuffle
@@ -716,8 +726,10 @@ define <16 x i16> @shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_2
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_16_00_16_00_16_00_16_08_24_08_24_08_24_08_24:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,0,1,u,u,0,1,u,u,0,1,u,u,0,1,u,u,16,17,u,u,16,17,u,u,16,17,u,u,16,17]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,3,4,5,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,2,0,4,5,6,7,8,8,10,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,4,6,4,8,9,10,11,12,12,14,12]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 0, i32 16, i32 0, i32 16, i32 0, i32 16, i32 8, i32 24, i32 8, i32 24, i32 8, i32 24, i32 8, i32 24>
@@ -740,7 +752,7 @@ define <16 x i16> @shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_1
 ;
 ; AVX2-LABEL: shuffle_v16i16_16_16_16_16_04_05_06_07_24_24_24_24_12_13_14_15:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,0,1,0,1,u,u,u,u,u,u,u,u,16,17,16,17,16,17,16,17,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 4, i32 5, i32 6, i32 7, i32 24, i32 24, i32 24, i32 24, i32 12, i32 13, i32 14, i32 15>
@@ -766,7 +778,8 @@ define <16 x i16> @shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_1
 ; AVX2-LABEL: shuffle_v16i16_19_18_17_16_07_06_05_04_27_26_25_24_15_14_13_12:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,4,5,2,3,0,1,14,15,12,13,10,11,8,9,22,23,20,21,18,19,16,17,30,31,28,29,26,27,24,25]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 19, i32 18, i32 17, i32 16, i32 7, i32 6, i32 5, i32 4, i32 27, i32 26, i32 25, i32 24, i32 15, i32 14, i32 13, i32 12>
   ret <16 x i16> %shuffle
@@ -787,8 +800,9 @@ define <16 x i16> @shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_19_18_17_16_03_02_01_00_27_26_25_24_11_10_09_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,6,7,4,5,2,3,0,1,u,u,u,u,u,u,u,u,22,23,20,21,18,19,16,17]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[6,7,4,5,2,3,0,1,u,u,u,u,u,u,u,u,22,23,20,21,18,19,16,17,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 19, i32 18, i32 17, i32 16, i32 3, i32 2, i32 1, i32 0, i32 27, i32 26, i32 25, i32 24, i32 11, i32 10, i32 9, i32 8>
@@ -807,7 +821,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_00_01_00_08_08_08_08_08_08_09_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,2,3,0,1,16,17,16,17,16,17,16,17,16,17,16,17,18,19,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,5,4,8,9,10,11,12,12,13,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 9, i32 8>
   ret <16 x i16> %shuffle
@@ -825,7 +841,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_00_02_00_00_08_08_08_08_08_10_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,4,5,0,1,0,1,16,17,16,17,16,17,16,17,16,17,20,21,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,6,4,4,8,9,10,11,12,14,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 2, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 8, i32 10, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -843,7 +861,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_03_00_00_00_08_08_08_08_11_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,6,7,0,1,0,1,0,1,16,17,16,17,16,17,16,17,22,23,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,1,4,5,4,5]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 3, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 8, i32 11, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -861,7 +881,9 @@ define <16 x i16> @shuffle_v16i16_00_00_00_04_00_00_00_00_08_08_08_12_08_08_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_04_00_00_00_00_08_08_08_12_08_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,8,9,0,1,0,1,0,1,0,1,16,17,16,17,16,17,24,25,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,0,3,4,6,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,2,4,5,6,7,8,8,8,10,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 4, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 8, i32 12, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -879,7 +901,9 @@ define <16 x i16> @shuffle_v16i16_00_00_05_00_00_00_00_00_08_08_13_08_08_08_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_05_00_00_00_00_00_08_08_13_08_08_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,10,11,0,1,0,1,0,1,0,1,0,1,16,17,16,17,26,27,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,0,3,4,6,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,3,0,4,5,6,7,8,8,11,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 5, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 8, i32 13, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -897,7 +921,9 @@ define <16 x i16> @shuffle_v16i16_00_06_00_00_00_00_00_00_08_14_08_08_08_08_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_06_00_00_00_00_00_00_08_14_08_08_08_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,12,13,0,1,0,1,0,1,0,1,0,1,0,1,16,17,28,29,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,3,0,3,4,7,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,0,0,4,5,6,7,8,10,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 6, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 8, i32 14, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -915,7 +941,9 @@ define <16 x i16> @shuffle_v16i16_07_00_00_00_00_00_00_00_15_08_08_08_08_08_08_0
 ;
 ; AVX2-LABEL: shuffle_v16i16_07_00_00_00_00_00_00_00_15_08_08_08_08_08_08_08:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,0,1,0,1,0,1,0,1,0,1,0,1,0,1,30,31,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,3,0,3,4,7,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 7, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 15, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
   ret <16 x i16> %shuffle
@@ -1236,7 +1264,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_2
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_16_16_16_16_20_20_20_20:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 16, i32 16, i32 16, i32 16, i32 20, i32 20, i32 20, i32 20>
   ret <16 x i16> %shuffle
@@ -1256,7 +1285,8 @@ define <16 x i16> @shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_2
 ; AVX2-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_16_16_16_16_20_20_20_20:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12, i32 16, i32 16, i32 16, i32 16, i32 20, i32 20, i32 20, i32 20>
   ret <16 x i16> %shuffle
@@ -1277,7 +1307,8 @@ define <16 x i16> @shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_2
 ; AVX2-LABEL: shuffle_v16i16_08_08_08_08_12_12_12_12_24_24_24_24_28_28_28_28:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 8, i32 8, i32 8, i32 8, i32 12, i32 12, i32 12, i32 12, i32 24, i32 24, i32 24, i32 24, i32 28, i32 28, i32 28, i32 28>
   ret <16 x i16> %shuffle
@@ -1297,7 +1328,8 @@ define <16 x i16> @shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_2
 ; AVX2-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_04_24_24_24_24_28_28_28_28:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,8,9,16,17,16,17,16,17,16,17,24,25,24,25,24,25,24,25]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 4, i32 4, i32 4, i32 24, i32 24, i32 24, i32 24, i32 28, i32 28, i32 28, i32 28>
   ret <16 x i16> %shuffle
@@ -2304,7 +2336,9 @@ define <16 x i16> @shuffle_v16i16_00_01_02_07_uu_uu_uu_uu_08_09_10_15_uu_uu_uu_u
 ;
 ; AVX2-LABEL: shuffle_v16i16_00_01_02_07_uu_uu_uu_uu_08_09_10_15_uu_uu_uu_uu:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,30,31,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,7,6,7,8,9,10,11,12,15,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 8, i32 9, i32 10, i32 15, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
@@ -2344,7 +2378,10 @@ define <16 x i16> @shuffle_v16i16_04_05_06_03_uu_uu_uu_uu_12_13_14_11_uu_uu_uu_u
 ;
 ; AVX2-LABEL: shuffle_v16i16_04_05_06_03_uu_uu_uu_uu_12_13_14_11_uu_uu_uu_uu:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,6,7,u,u,u,u,u,u,u,u,24,25,26,27,28,29,22,23,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,3,2,3,4,5,6,7,8,11,10,11,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[2,3,0,1,4,5,6,7,10,11,8,9,12,13,14,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 5, i32 6, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 12, i32 13, i32 14, i32 11, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
@@ -2554,7 +2591,8 @@ define <16 x i16> @shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_3
 ; AVX2-NEXT:    vpshuflw {{.*#+}} xmm2 = xmm2[0,0,2,1,4,5,6,7]
 ; AVX2-NEXT:    vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
 ; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,u,u,2,3,u,u,12,13,u,u,14,15,u,u,16,17,u,u,18,19,u,u,28,29,u,u,30,31,u,u]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 6, i32 22, i32 7, i32 31, i32 8, i32 24, i32 9, i32 25, i32 14, i32 30, i32 15, i32 31>
@@ -2584,7 +2622,8 @@ define <16 x i16> @shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_2
 ; AVX2-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5,6,7]
 ; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX2-NEXT:    vinserti128 $1, %xmm4, %ymm1, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,u,u,2,3,u,u,12,13,u,u,14,15,u,u,16,17,u,u,18,19,u,u,28,29,u,u,30,31,u,u]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 20, i32 1, i32 21, i32 6, i32 16, i32 7, i32 25, i32 8, i32 28, i32 9, i32 29, i32 14, i32 24, i32 15, i32 25>
@@ -2614,7 +2653,9 @@ define <16 x i16> @shuffle_v16i16_01_00_17_16_03_02_19_26_09_08_25_24_11_10_27_2
 ; AVX2-NEXT:    vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
 ; AVX2-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,2,3,0,1,8,9,10,11,6,7,4,5]
 ; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,0,1,u,u,u,u,6,7,4,5,u,u,u,u,18,19,16,17,u,u,u,u,22,23,20,21,u,u,u,u]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,2,1,4,5,6,5]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,0,2,3,4,5,6,7,9,8,10,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,6,7,8,9,10,11,15,14,14,15]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 1, i32 0, i32 17, i32 16, i32 3, i32 2, i32 19, i32 26, i32 9, i32 8, i32 25, i32 24, i32 11, i32 10, i32 27, i32 26>
@@ -2723,7 +2764,8 @@ define <16 x i16> @shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_u
 ; AVX2-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7,8,9],ymm1[10],ymm0[11,12,13,14,15]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,8,9,6,7,4,5,u,u,u,u,u,u,u,u,24,25,24,25,22,23,20,21,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,1,2,3,6,5,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,3,2,4,5,6,7,8,8,11,10,12,13,14,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 4, i32 3, i32 18, i32 undef, i32 undef, i32 undef, i32 undef, i32 12, i32 12, i32 11, i32 26, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
@@ -2745,7 +2787,10 @@ define <16 x i16> @shuffle_v16i16_00_03_02_21_uu_uu_uu_uu_08_11_10_29_uu_uu_uu_u
 ; AVX2-LABEL: shuffle_v16i16_00_03_02_21_uu_uu_uu_uu_08_11_10_29_uu_uu_uu_uu:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,6,7,4,5,10,11,u,u,u,u,u,u,u,u,16,17,22,23,20,21,26,27,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,6,7,8,9,10,11,14,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,1,2,3,6,5,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,3,2,1,4,5,6,7,8,11,10,9,12,13,14,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 3, i32 2, i32 21, i32 undef, i32 undef, i32 undef, i32 undef, i32 8, i32 11, i32 10, i32 29, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i16> %shuffle
@@ -2893,7 +2938,8 @@ define <16 x i16> @shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_2
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm2
 ; AVX2-NEXT:    vpermd %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,2,3,u,u,10,11,14,15,u,u,u,u,u,u,u,u,18,19,u,u,26,27,30,31,u,u]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,2,1,4,5,6,7,8,9,10,9,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,7,7,8,9,10,11,12,13,15,15]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 undef, i32 undef, i32 undef, i32 1, i32 undef, i32 5, i32 7, i32 25, i32 undef, i32 undef, i32 undef, i32 9, i32 undef, i32 13, i32 15, i32 25>
@@ -2917,8 +2963,10 @@ define <16 x i16> @shuffle_v16i16_uu_uu_04_uu_16_18_20_uu_uu_uu_12_uu_24_26_28_u
 ;
 ; AVX2-LABEL: shuffle_v16i16_uu_uu_04_uu_16_18_20_uu_uu_uu_12_uu_24_26_28_uu:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,0,1,4,5,8,9,u,u,u,u,u,u,u,u,u,u,16,17,20,21,24,25,u,u]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,2,2,3,4,5,6,7,8,10,10,11,12,13,14,15]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,6,7,4,7,8,9,10,11,14,15,12,15]
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 undef, i32 undef, i32 4, i32 undef, i32 16, i32 18, i32 20, i32 undef, i32 undef, i32 undef, i32 12, i32 undef, i32 24, i32 26, i32 28, i32 undef>
@@ -3230,7 +3278,9 @@ define <16 x i16> @shuffle_v16i16_23_uu_03_uu_20_20_05_uu_31_uu_11_uu_28_28_13_u
 ; AVX2-LABEL: shuffle_v16i16_23_uu_03_uu_20_20_05_uu_31_uu_11_uu_28_28_13_uu:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7],ymm0[8,9,10,11],ymm1[12],ymm0[13,14],ymm1[15]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,u,u,6,7,u,u,8,9,8,9,10,11,u,u,30,31,u,u,22,23,u,u,24,25,24,25,26,27,u,u]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,1,2,3,7,5,6,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,1,3,3,4,5,6,7,9,9,11,11,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,5,7,8,9,10,11,12,12,13,15]
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 23, i32 undef, i32 3, i32 undef, i32 20, i32 20, i32 5, i32 undef, i32 31, i32 undef, i32 11, i32 undef, i32 28, i32 28, i32 13, i32 undef>
   ret <16 x i16> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll
index ed3c66637eddedc5e1a82e66b4e1e28f883bc8af..d1a762e7165d30becbde75ea3568828073cf0ef6 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -985,7 +985,9 @@ define <32 x i8> @shuffle_v32i8_00_32_00_32_00_32_00_32_00_32_00_32_00_32_00_32_
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpxor %ymm2, %ymm2, %ymm2
 ; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,16,17,16,17,16,17,16,17,16,17,16,17,16,17,16,17]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,0,3,4,5,4,7]
+; AVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
+; AVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
 ; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
 ; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq