[X86] Improve AVX1 shuffle lowering for v8f32 shuffles where the low half comes from V1 and the high half comes from V2 and the halves do the same operation

[craig.topper]

Summary

To lower this we now create a new V1 containing the low half of both sources and a new V2 containing the upper half of both sources. Then we created a repeated lane shuffle of those new sources to create the final result.

This fixes PR35833

[kbelochapka]

currently you call lowerVectorShuffleSplitLowHigh() form lowerV8F32VectorShuffle() only.

In order to cover v4f64, v4i64, v8i32 types,
can you add a call to lowerVectorShuffleSplitLowHigh() to other 256 bit vector lowering functions?
lowerV4F64VectorShuffle()
lowerV4I64VectorShuffle()
lowerV8I32VectorShuffle()

[kbelochapka]

can you add a test for "128 bit line swapped" shuffle masks: [8, 10, 12, 14, 0, 2, 4, 6,], [9, 11, 13, 15, 1, 3, 5, 7 ]

[craig.topper]

Add support for v4f64 as well.

Without AVX2, integer types are cast to FP so this supports v8i32/v4i64 without AVX2 as is.

[RKSimon]

What is the point of doing the adds/fadds in the new tests? It looks like its just to test multiple shuffle per test or is there a domain aspect to it that I'm missing?

[craig.topper]

The adds/fadds are just there cause that's what was in the bug report. And I wanted to make sure we were able to reuse the vperm2f128 shuffles that create NewV1 and NewV2.

[kbelochapka]

Hi Craig, I placed simple test on the Bugzilla, your fix seems to be working fine.
The only thing, there are couple or more of shuffle masks which looks like they can be a subject of the similar optimization, but the generated code for them remained not optimized.
IN0: |0|1| | |4|5| | |
IN1: |8|9| | | | |E|F|
MASK011: 0,1,4,5,8,9,14,15
IN0: |0|1| | | | |6|7|
IN1: |8|9| | |C|D| | |
MASK012: 0,1,6,7,8,9,12,13

[craig.topper]

Can you show what code you expect for those cases and I'll see what I can do?

[kbelochapka]

Four instructions perhaps:

vextractf128, vextractf128, vshufps, vblendps?

It is better than current six instructions.

[RKSimon]

In D41794#969440, @craig.topper wrote:

The adds/fadds are just there cause that's what was in the bug report. And I wanted to make sure we were able to reuse the vperm2f128 shuffles that create NewV1 and NewV2.

OK - please can you commit the tests to trunk so the patch shows the codegen diff?

[craig.topper]

Rebase now that tets have been committed.

[craig.topper]

Actually rebase the patch.

[RKSimon]

This looks like there is a lot of crossover with lowerVectorShuffleByMerging128BitLanes?

lib/Target/X86/X86ISelLowering.cpp

10761

Should't RepeatMask be just LaneSize wide?

10791

Do we gain anything by relaxing this and keeping PermuteMask[i] as UNDEF if the original Mask[i] was UNDEF?

[craig.topper]

bYeah there may be some crossover with lowerVectorShuffleByMerging128BitLanes. I'll see if I can generalize lowerVectorShuffleByMerging128BitLanes more.

lib/Target/X86/X86ISelLowering.cpp

10761

Yes it should.

10791

Not sure. I was trying to create a repeated lane shuffle so its based on both lanes. If its undef in both lanes it will be undef here.

[RKSimon]

@craig.topper Are you still looking at this?

[craig.topper]

Replace the contents of lowerVectorShuffleByMerging128BitLanes. Now we try to form 2 lane correcting shuffles and a repeated shuffle mask.

The code goes through 2 passes to determine the repeated mask. First if looks for any lanes that need 2 source lanes and build a repeating mask from that if any exist. Then it will try to fit one source lanes to that mask duplicating the same source lane on both inputs if necessary to match the repeated mask.

This gets the case in PR36933 now too. I haven't added the test for it to the patch yet.

It's regressing some of the avx512 cases now though because we don't seem to be able to combine 2 source permi2q with a perm2f128 feeding it.

[craig.topper]

I forget to mention, I tried to write the code in a way that it should be somewhat straightforward to extend to 512 bits with 128 bit lanes.

[craig.topper]

Looks like X86ISD::SHUF128 isn't recognized as a target shuffle. That explains some of the regressions here. I'll submit a separate patch for that and we can rebase this one.

[RKSimon]

Thanks @craig.topper - please can you update the diff with context when you rebase? It can wait until the X86ISD::SHUF128 fix

[craig.topper]

Rebase and full context

[RKSimon]

lib/Target/X86/X86ISelLowering.cpp

13232

There is a is128BitLaneRepeatedShuffleMask(VT, Mask) version that you can use instead

[RKSimon]

Some minors and I'm a bit concerned about the regression in prefer-avx256-mask-shuffle.ll - not sure why it regresses so much.

lib/Target/X86/X86ISelLowering.cpp

13224

This comment + FIXME needs updating

13269

This seems easier to grok (at least to me):

int Src;
if (Srcs[0] < 0 || Srcs[0] == LaneSrc)
  Src = 0;
else if (Srcs[1] < 0 || Srcs[1] == LaneSrc)
  Src = 1;
else
  return SDValue();

Srcs[Src] = LaneSrc;
InLaneMask[i] = (M % LaneSize) + Src * Size;

13281

for (int i = 0, e = M1.size(); i != e; ++i)

13289

for (int i = 0, e = MergedMask.size(); i != e; ++i) {

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

210

This is the only notable regression - any idea why it breaks so badly?

[craig.topper]

Address review comments. Will investigate the test regression

[craig.topper]

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

210

It looks like we go through lowerVectorShuffleAsLanePermuteAndBlend which makes the unary shuffle non-unary. Then we go through lowerVectorShuffleByMerging128BitLanes which creates a repeated mask. But we still weren't able to handle this repeated mask cleanly so we end up shuffling and blending.

[craig.topper]

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

210

Why can't shuffle combining merge the two vblendds with the vpermq to create two new vpermqs? Is it because the vpermq is used twice or the fact that vblendd is v8i32 and vpermq is v4i64? Or something else?

[RKSimon]

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

210

Target shuffle combining mainly combines to single ops only, there might be scope to improve this if lower1BitVectorShuffle avoids mask vector extension to 512-bit when prefer-256 is enabled.

[craig.topper]

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

210

The blends are on separate dependency chains but the each use the same vpermq. But I think each blend could independently be replaced with a vpermq removing the shared vpermq

[craig.topper]

Rebase

[RKSimon]

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

210

This seems to be the only remaining issue - are we better off commiting now or finding a fix first?

[craig.topper]

I'm fine with committing it if you are.

[RKSimon]

LGTM - I'll continue investigating how to combine VSELECT with target shuffles to fix that prefer-avx256-mask-shuffle.ll regression.

[RKSimon]

Committed at rL339818 (typo in commit message prevented phab from catching it)

lib/Target/X86/X86ISelLowering.cpp

     const X86Subtarget &Subtarget, SelectionDAG &DAG) {
   assert(!V2.isUndef() && "This is only useful with multiple inputs.");
 
+  {
+    SmallVector<int, 8> RepeatedShuffleMask;
+    if (is128BitLaneRepeatedShuffleMask(VT, Mask, RepeatedShuffleMask))
+      return SDValue();
+  }
+
   int Size = Mask.size();
   int LaneSize = 128 / VT.getScalarSizeInBits();
   int NumLanes = Size / LaneSize;
-  assert(NumLanes > 1 && "Only handles 256-bit and wider shuffles.");
+  assert(NumLanes == 2 && "Only handles 256-bit shuffles.");
+
+  SmallVector<int, 16> RepeatMask(LaneSize, -1);
+  int LaneSrcs[2][2] = { { -1, -1 }, { -1 , -1 } };
+
+  // First pass will try to fill in the RepeatMask from lanes that need two
+  // sources.
+  for (int Lane = 0; Lane != NumLanes; ++Lane) {
+    int Srcs[2] = { -1, -1 };
+    SmallVector<int, 16> InLaneMask(LaneSize, -1);
+    for (int i = 0; i != LaneSize; ++i) {
+      int M = Mask[(Lane * LaneSize) + i];
+      if (M < 0)
+        continue;
+      // Determine which of the 4 possible input lanes (2 from each source)
+      // this element comes from. Assign that as one of the sources for this
+      // lane. We can assign up to 2 sources for this lane. If we run out
+      // sources we can't do anything.
+      int LaneSrc = M / LaneSize;
+      int Src = 0;
+      if (Srcs[0] < 0)
+        Srcs[0] = LaneSrc;
+      if (Srcs[0] != LaneSrc) {
+        Src = 1;
+        if (Srcs[1] < 0)
+          Srcs[1] = LaneSrc;
+        if (Srcs[1] != LaneSrc)
+          return SDValue();
+      }
 
-  // See if we can build a hypothetical 128-bit lane-fixing shuffle mask. Also
-  // check whether the in-128-bit lane shuffles share a repeating pattern.
-  SmallVector<int, 4> Lanes((unsigned)NumLanes, -1);
-  SmallVector<int, 4> InLaneMask((unsigned)LaneSize, -1);
-  for (int i = 0; i < Size; ++i) {
-    if (Mask[i] < 0)
+      InLaneMask[i] = (M % LaneSize) + Src * Size;
+    }
+
+    // If this lane has two sources, see if it fits with the repeat mask so far.
+    if (Srcs[1] < 0)
       continue;
 
-    int j = i / LaneSize;
+    LaneSrcs[Lane][0] = Srcs[0];
+    LaneSrcs[Lane][1] = Srcs[1];
 
-    if (Lanes[j] < 0) {
-      // First entry we've seen for this lane.
-      Lanes[j] = Mask[i] / LaneSize;
-    } else if (Lanes[j] != Mask[i] / LaneSize) {
-      // This doesn't match the lane selected previously!
-      return SDValue();
+    auto MatchMasks = [](ArrayRef<int> M1, ArrayRef<int> M2) {
+      assert(M1.size() == M2.size() && "Unexpected mask size");
+      for (int i = 0; i != (int)M1.size(); ++i)
+        if (M1[i] >= 0 && M2[i] >= 0 && M1[i] != M2[i])
+          return false;
+      return true;
+    };
+
+    auto MergeMasks = [](ArrayRef<int> Mask, MutableArrayRef<int> MergedMask) {
+      assert(Mask.size() == MergedMask.size() && "Unexpected mask size");
+      for (int i = 0; i != (int)MergedMask.size(); ++i) {
+        int M = Mask[i];
+        if (M < 0)
+          continue;
+        assert((MergedMask[i] < 0 || MergedMask[i] == M) &&
+               "Unexpected mask element");
+        MergedMask[i] = M;
+      }
+    };
+
+    if (MatchMasks(InLaneMask, RepeatMask)) {
+      // Merge this lane mask into the final repeat mask.
+      MergeMasks(InLaneMask, RepeatMask);
+      continue;
     }
 
-    // Check that within each lane we have a consistent shuffle mask.
-    int k = i % LaneSize;
-    if (InLaneMask[k] < 0) {
-      InLaneMask[k] = Mask[i] % LaneSize;
-    } else if (InLaneMask[k] != Mask[i] % LaneSize) {
-      // This doesn't fit a repeating in-lane mask.
-      return SDValue();
+    // Didn't find a match. Swap the operands and try again.
+    std::swap(LaneSrcs[Lane][0], LaneSrcs[Lane][1]);
+    ShuffleVectorSDNode::commuteMask(InLaneMask);
+
+    if (MatchMasks(InLaneMask, RepeatMask)) {
+      // Merge this lane mask into the final repeat mask.
+      MergeMasks(InLaneMask, RepeatMask);
+      continue;
     }
+
+    // Couldn't find a match with the operands in either order.
+    return SDValue();
   }
 
-  // First shuffle the lanes into place.
-  MVT LaneVT = MVT::getVectorVT(VT.isFloatingPoint() ? MVT::f64 : MVT::i64,
-                                VT.getSizeInBits() / 64);
-  SmallVector<int, 8> LaneMask((unsigned)NumLanes * 2, -1);
-  for (int i = 0; i < NumLanes; ++i)
-    if (Lanes[i] >= 0) {
-      LaneMask[2 * i + 0] = 2*Lanes[i] + 0;
-      LaneMask[2 * i + 1] = 2*Lanes[i] + 1;
+  // Now handle any lanes with only one source.
+  for (int Lane = 0; Lane != NumLanes; ++Lane) {
+    // If this lane has already been processed, skip it.
+    if (LaneSrcs[Lane][0] >= 0)
+      continue;
+
+    for (int i = 0; i != LaneSize; ++i) {
+      int M = Mask[(Lane * LaneSize) + i];
+      if (M < 0)
+        continue;
+
+      // If RepeatMask isn't defined yet we can define it ourself.
+      if (RepeatMask[i] < 0)
+        RepeatMask[i] = M % LaneSize;
+
+      if (RepeatMask[i] < Size) {
+        if (RepeatMask[i] != M % LaneSize)
+          return SDValue();
+        LaneSrcs[Lane][0] = M / LaneSize;
+      } else {
+        if (RepeatMask[i] != ((M % LaneSize) + Size))
+          return SDValue();
+        LaneSrcs[Lane][1] = M / LaneSize;
+      }
     }
 
-  V1 = DAG.getBitcast(LaneVT, V1);
-  V2 = DAG.getBitcast(LaneVT, V2);
-  SDValue LaneShuffle = DAG.getVectorShuffle(LaneVT, DL, V1, V2, LaneMask);
+    if (LaneSrcs[Lane][0] < 0 && LaneSrcs[Lane][1] < 0)
+      return SDValue();
+  }
 
-  // Cast it back to the type we actually want.
-  LaneShuffle = DAG.getBitcast(VT, LaneShuffle);
+  SmallVector<int, 16> NewMask(Size, -1);
+  for (int Lane = 0; Lane != NumLanes; ++Lane) {
+    int Src = LaneSrcs[Lane][0];
+    for (int i = 0; i != LaneSize; ++i) {
+      int M = -1;
+      if (Src >= 0)
+        M = Src * LaneSize + i;
+      NewMask[Lane * LaneSize + i] = M;
+    }
+  }
+  SDValue NewV1 = DAG.getVectorShuffle(VT, DL, V1, V2, NewMask);
 
-  // Now do a simple shuffle that isn't lane crossing.
-  SmallVector<int, 8> NewMask((unsigned)Size, -1);
-  for (int i = 0; i < Size; ++i)
-    if (Mask[i] >= 0)
-      NewMask[i] = (i / LaneSize) * LaneSize + Mask[i] % LaneSize;
-  assert(!is128BitLaneCrossingShuffleMask(VT, NewMask) &&
-         "Must not introduce lane crosses at this point!");
+  for (int Lane = 0; Lane != NumLanes; ++Lane) {
+    int Src = LaneSrcs[Lane][1];
+    for (int i = 0; i != LaneSize; ++i) {
+      int M = -1;
+      if (Src >= 0)
+        M = Src * LaneSize + i;
+      NewMask[Lane * LaneSize + i] = M;
+    }
+  }
+  SDValue NewV2 = DAG.getVectorShuffle(VT, DL, V1, V2, NewMask);
 
-  return DAG.getVectorShuffle(VT, DL, LaneShuffle, DAG.getUNDEF(VT), NewMask);
+  for (int i = 0; i != Size; ++i) {
+    NewMask[i] = RepeatMask[i % LaneSize];
+    if (NewMask[i] < 0)
+      continue;
+
+    NewMask[i] += (i / LaneSize) * LaneSize;
+  }
+  return DAG.getVectorShuffle(VT, DL, NewV1, NewV2, NewMask);
 }
 
 /// Lower shuffles where an entire half of a 256 or 512-bit vector is UNDEF.

test/CodeGen/X86/avx512-shuffles/partial_permute.ll

 ; CHECK-LABEL: test_8xi64_to_4xi64_perm_mem_mask6:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm1
-; CHECK-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; CHECK-NEXT:    vmovdqa {{.*#+}} ymm0 = [7,2,3,2]
-; CHECK-NEXT:    vpermi2q %ymm2, %ymm1, %ymm0
+; CHECK-NEXT:    vextracti64x4 $1, %zmm1, %ymm0
+; CHECK-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
+; CHECK-NEXT:    vmovdqa {{.*#+}} ymm0 = [1,6,3,6]
+; CHECK-NEXT:    vpermi2q %ymm1, %ymm2, %ymm0
 ; CHECK-NEXT:    retq
   %vec = load <8 x i64>, <8 x i64>* %vp
   %res = shufflevector <8 x i64> %vec, <8 x i64> undef, <4 x i32> <i32 7, i32 2, i32 3, i32 2>

 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vextracti64x4 $1, %zmm2, %ymm3
-; CHECK-NEXT:    vmovdqa {{.*#+}} ymm4 = [7,2,3,2]
-; CHECK-NEXT:    vpermi2q %ymm3, %ymm2, %ymm4
+; CHECK-NEXT:    vperm2i128 {{.*#+}} ymm3 = ymm3[2,3],ymm2[2,3]
+; CHECK-NEXT:    vmovdqa {{.*#+}} ymm4 = [1,6,3,6]
+; CHECK-NEXT:    vpermi2q %ymm2, %ymm3, %ymm4
 ; CHECK-NEXT:    vptestnmq %ymm1, %ymm1, %k1
 ; CHECK-NEXT:    vmovdqa64 %ymm4, %ymm0 {%k1}
 ; CHECK-NEXT:    retq

 ; CHECK-LABEL: test_masked_z_8xi64_to_4xi64_perm_mem_mask6:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
-; CHECK-NEXT:    vextracti64x4 $1, %zmm2, %ymm3
-; CHECK-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,2,3,2]
+; CHECK-NEXT:    vextracti64x4 $1, %zmm2, %ymm1
+; CHECK-NEXT:    vperm2i128 {{.*#+}} ymm3 = ymm1[2,3],ymm2[2,3]
+; CHECK-NEXT:    vmovdqa {{.*#+}} ymm1 = [1,6,3,6]
 ; CHECK-NEXT:    vptestnmq %ymm0, %ymm0, %k1
-; CHECK-NEXT:    vpermi2q %ymm3, %ymm2, %ymm1 {%k1} {z}
+; CHECK-NEXT:    vpermi2q %ymm2, %ymm3, %ymm1 {%k1} {z}
 ; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
 ; CHECK-NEXT:    retq
   %vec = load <8 x i64>, <8 x i64>* %vp

 define <4 x double> @test_8xdouble_to_4xdouble_perm_mem_mask0(<8 x double>* %vp) {
 ; CHECK-LABEL: test_8xdouble_to_4xdouble_perm_mem_mask0:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vmovapd (%rdi), %zmm1
-; CHECK-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
-; CHECK-NEXT:    vmovapd {{.*#+}} ymm0 = [1,6,7,2]
-; CHECK-NEXT:    vpermi2pd %ymm2, %ymm1, %ymm0
+; CHECK-NEXT:    vmovapd (%rdi), %zmm0
+; CHECK-NEXT:    vextractf64x4 $1, %zmm0, %ymm1
+; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
+; CHECK-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; CHECK-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[0],ymm0[3],ymm2[2]
 ; CHECK-NEXT:    retq
   %vec = load <8 x double>, <8 x double>* %vp
   %res = shufflevector <8 x double> %vec, <8 x double> undef, <4 x i32> <i32 1, i32 6, i32 7, i32 2>

 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm2
 ; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm3
-; CHECK-NEXT:    vmovapd {{.*#+}} ymm4 = [1,6,7,2]
-; CHECK-NEXT:    vpermi2pd %ymm3, %ymm2, %ymm4
-; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; CHECK-NEXT:    vcmpeqpd %ymm2, %ymm1, %k1
-; CHECK-NEXT:    vmovapd %ymm4, %ymm0 {%k1}
+; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],ymm2[2,3]
+; CHECK-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3]
+; CHECK-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; CHECK-NEXT:    vcmpeqpd %ymm3, %ymm1, %k1
+; CHECK-NEXT:    vshufpd {{.*#+}} ymm0 {%k1} = ymm2[1],ymm4[0],ymm2[3],ymm4[2]
 ; CHECK-NEXT:    retq
   %vec = load <8 x double>, <8 x double>* %vp
   %shuf = shufflevector <8 x double> %vec, <8 x double> undef, <4 x i32> <i32 1, i32 6, i32 7, i32 2>

 define <4 x double> @test_masked_z_8xdouble_to_4xdouble_perm_mem_mask0(<8 x double>* %vp, <4 x double> %mask) {
 ; CHECK-LABEL: test_masked_z_8xdouble_to_4xdouble_perm_mem_mask0:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vmovapd (%rdi), %zmm2
-; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm3
-; CHECK-NEXT:    vmovapd {{.*#+}} ymm1 = [1,6,7,2]
-; CHECK-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; CHECK-NEXT:    vcmpeqpd %ymm4, %ymm0, %k1
-; CHECK-NEXT:    vpermi2pd %ymm3, %ymm2, %ymm1 {%k1} {z}
-; CHECK-NEXT:    vmovapd %ymm1, %ymm0
+; CHECK-NEXT:    vmovapd (%rdi), %zmm1
+; CHECK-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
+; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm2[2,3],ymm1[2,3]
+; CHECK-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; CHECK-NEXT:    vcmpeqpd %ymm2, %ymm0, %k1
+; CHECK-NEXT:    vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm1[1],ymm3[0],ymm1[3],ymm3[2]
 ; CHECK-NEXT:    retq
   %vec = load <8 x double>, <8 x double>* %vp
   %shuf = shufflevector <8 x double> %vec, <8 x double> undef, <4 x i32> <i32 1, i32 6, i32 7, i32 2>

 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm2
 ; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm3
-; CHECK-NEXT:    vmovapd {{.*#+}} ymm4 = [1,2,3,4]
-; CHECK-NEXT:    vpermi2pd %ymm3, %ymm2, %ymm4
-; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; CHECK-NEXT:    vcmpeqpd %ymm2, %ymm1, %k1
-; CHECK-NEXT:    vmovapd %ymm4, %ymm0 {%k1}
+; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm2[2,3],ymm3[0,1]
+; CHECK-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; CHECK-NEXT:    vcmpeqpd %ymm4, %ymm1, %k1
+; CHECK-NEXT:    vshufpd {{.*#+}} ymm0 {%k1} = ymm2[1],ymm3[0],ymm2[3],ymm3[2]
 ; CHECK-NEXT:    retq
   %vec = load <8 x double>, <8 x double>* %vp
   %shuf = shufflevector <8 x double> %vec, <8 x double> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 4>

 define <4 x double> @test_masked_z_8xdouble_to_4xdouble_perm_mem_mask2(<8 x double>* %vp, <4 x double> %mask) {
 ; CHECK-LABEL: test_masked_z_8xdouble_to_4xdouble_perm_mem_mask2:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vmovapd (%rdi), %zmm2
-; CHECK-NEXT:    vextractf64x4 $1, %zmm2, %ymm3
-; CHECK-NEXT:    vmovapd {{.*#+}} ymm1 = [1,2,3,4]
-; CHECK-NEXT:    vpxor %xmm4, %xmm4, %xmm4
-; CHECK-NEXT:    vcmpeqpd %ymm4, %ymm0, %k1
-; CHECK-NEXT:    vpermi2pd %ymm3, %ymm2, %ymm1 {%k1} {z}
-; CHECK-NEXT:    vmovapd %ymm1, %ymm0
+; CHECK-NEXT:    vmovapd (%rdi), %zmm1
+; CHECK-NEXT:    vextractf64x4 $1, %zmm1, %ymm2
+; CHECK-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm2[0,1]
+; CHECK-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; CHECK-NEXT:    vcmpeqpd %ymm3, %ymm0, %k1
+; CHECK-NEXT:    vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm1[1],ymm2[0],ymm1[3],ymm2[2]
 ; CHECK-NEXT:    retq
   %vec = load <8 x double>, <8 x double>* %vp
   %shuf = shufflevector <8 x double> %vec, <8 x double> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 4>

test/CodeGen/X86/prefer-avx256-mask-shuffle.ll

 ; AVX256VLBW-NEXT:    vptestnmb %ymm0, %ymm0, %k0
 ; AVX256VLBW-NEXT:    vpmovm2b %k0, %ymm0
 ; AVX256VLBW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX256VLBW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[3,6,u,12,3,7,7,0,3,6,1,13,3,u,7,0,u,u,22,u,u,u,u,u,u,u,u,u,u,21,u,u]
-; AVX256VLBW-NEXT:    movl $-537190396, %eax # imm = 0xDFFB2004
+; AVX256VLBW-NEXT:    vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX256VLBW-NEXT:    vpshufd {{.*#+}} ymm2 = ymm2[1,1,2,1,5,5,6,5]
+; AVX256VLBW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX256VLBW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[3,6,u,12,3,7,7,0,3,6,1,13,3,u,7,0,19,22,u,28,19,23,23,16,19,22,17,29,19,u,23,16]
+; AVX256VLBW-NEXT:    movl $537141252, %eax # imm = 0x20042004
 ; AVX256VLBW-NEXT:    kmovd %eax, %k1
-; AVX256VLBW-NEXT:    vpshufb {{.*#+}} ymm0 {%k1} = ymm1[u,u,6,u,u,u,u,u,u,u,u,u,u,5,u,u,19,22,u,28,19,23,23,16,19,22,17,29,19,u,23,16]
+; AVX256VLBW-NEXT:    vmovdqu8 %ymm2, %ymm0 {%k1}
 ; AVX256VLBW-NEXT:    vpmovb2m %ymm0, %k0
 ; AVX256VLBW-NEXT:    vpmovm2b %k0, %ymm0
 ; AVX256VLBW-NEXT:    retq

test/CodeGen/X86/vector-shuffle-256-v16.ll

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpbroadcastw %xmm0, %xmm1
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,2,3,4,5,6,7,8,8,10,11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-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-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpslldq {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,17]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpbroadcastw %xmm0, %xmm1
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-FAST-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-FAST-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-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpslldq {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,17]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_00_00_00_00_00_00_00_00:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_01_00_01_02_03_02_11_08_09_08_09_10_11_10_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastq %xmm1, %xmm1
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,2,2,2]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,1,1,4,4,5,5]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; 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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_01_00_01_02_03_02_11_08_09_08_09_10_11_10_11:

 ;
 ; AVX2-LABEL: shuffle_v16i16_06_07_04_05_02_03_00_09_14_15_12_13_10_11_08_09:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7,8],ymm1[9],ymm0[10,11,12,13,14,15]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_04_05_06_07_16_17_18_27_12_13_14_15_24_25_26_27:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vpbroadcastw %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpslldq {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16,17]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,2,3,4,5,6,7,8,8,10,11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-FAST-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-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = 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-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08:

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vpsllq $48, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpsllq $48, %ymm1, %ymm1
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,0,0,0,4,5,6,7,8,8,8,8,12,13,14,15]
-; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,4,8,9,10,11,12,12,12,12]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,4,4,7,8,9,10,11,12,12,12,15]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-FAST-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-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,0,1,0,1,0,1,8,9,8,9,8,9,14,15,16,17,16,17,16,17,16,17,24,25,24,25,24,25,30,31]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_00_00_04_04_04_12_08_08_08_08_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_uu_00_uu_01_uu_02_uu_11_uu_08_uu_09_uu_10_uu_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,255,255,255,255,0,0,u,u,u,u,u,u,u,u,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_uu_04_uu_05_uu_06_uu_15_uu_12_uu_13_uu_14_uu_15:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm1 = ymm0[4,4,5,5,6,6,7,7,12,12,13,13,14,14,15,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_uu_04_uu_05_uu_06_uu_15_uu_12_uu_13_uu_14_uu_15:

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7,8,9,10,11,12],ymm1[13],ymm0[14,15]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[3,1,2,0,4,5,6,7,11,9,10,8,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_03_01_02_00_06_07_04_13_11_09_10_08_14_15_12_13:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7,8,9,10,11,12],ymm1[13],ymm0[14,15]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,2,3,4,5,0,1,12,13,14,15,8,9,10,11,22,23,18,19,20,21,16,17,28,29,30,31,24,25,26,27]
 ; AVX2-FAST-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_04_04_04_04_00_00_00_08_12_12_12_12_08_08_08_08:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,8,9,8,9,8,9,0,1,0,1,0,1,0,1,24,25,24,25,24,25,24,25,16,17,16,17,16,17,16,17]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[8,9,8,9,8,9,8,9,0,1,0,1,0,1,14,15,24,25,24,25,24,25,24,25,16,17,16,17,16,17,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_04_04_04_00_00_00_08_12_12_12_12_08_08_08_08:

 ;
 ; AVX2-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_13_10_11_08_09_14_15_12_13:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7,8,9,10,11,12],ymm1[13],ymm0[14,15]
 ; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,u,u,255,255,255,255,255,255,0,0,255,255,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255>
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7,8,9,10,11,12],ymm1[13],ymm0[14,15]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[2,3,0,2,4,5,6,7,10,11,8,10,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,1,3,2,4,5,7,6]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_02_03_00_02_06_07_04_13_10_11_08_10_14_15_12_13:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,u,u,255,255,255,255,255,255,0,0,255,255,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255>
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7,8,9,10,11,12],ymm1[13],ymm0[14,15]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,6,7,0,1,4,5,12,13,14,15,8,9,10,11,20,21,22,23,16,17,20,21,28,29,30,31,24,25,26,27]
 ; AVX2-FAST-NEXT:    retq
 ;

 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-SLOW-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
-; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,7,4,7,8,9,10,11,14,15,12,15]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
-; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[4,5,6,7,0,1,2,3,12,13,14,15,8,9,14,15,20,21,22,23,16,17,18,19,28,29,30,31,24,25,30,31]
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    retq
+; AVX2-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
+; 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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_02_03_00_01_06_07_04_15_10_11_08_09_14_15_12_15:
 ; AVX512VL:       # %bb.0:

 ;
 ; AVX2-LABEL: shuffle_v16i16_07_05_06_04_03_01_02_08_15_13_14_12_11_09_10_08:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[14,15,10,11,12,13,8,9,6,7,2,3,4,5,0,1,30,31,26,27,28,29,24,25,22,23,18,19,20,21,16,17]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_01_00_05_04_05_04_01_08_09_08_13_12_13_12_09_08:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,0,1,10,11,8,9,10,11,8,9,2,3,0,1,18,19,16,17,26,27,24,25,26,27,24,25,18,19,16,17]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[2,3,0,1,10,11,8,9,10,11,8,9,2,3,2,3,18,19,16,17,26,27,24,25,26,27,24,25,18,19,18,19]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_01_00_05_04_05_04_01_08_09_08_13_12_13_12_09_08:

 ;
 ; AVX2-LABEL: shuffle_v16i16_05_04_01_00_05_04_01_08_13_12_09_08_13_12_09_08:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[10,11,8,9,2,3,0,1,10,11,8,9,2,3,0,1,26,27,24,25,18,19,16,17,26,27,24,25,18,19,16,17]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[10,11,8,9,2,3,0,1,10,11,8,9,2,3,2,3,26,27,24,25,18,19,16,17,26,27,24,25,18,19,18,19]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_05_04_01_00_05_04_01_08_13_12_09_08_13_12_09_08:

 ;
 ; AVX2-LABEL: shuffle_v16i16_05_04_01_00_01_00_05_12_13_12_09_08_09_08_13_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[10,11,8,9,2,3,0,1,2,3,0,1,10,11,8,9,26,27,24,25,18,19,16,17,18,19,16,17,26,27,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[10,11,8,9,2,3,0,1,2,3,0,1,10,11,2,3,26,27,24,25,18,19,16,17,18,19,16,17,26,27,18,19]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_05_04_01_00_01_00_05_12_13_12_09_08_09_08_13_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_04_04_00_00_04_04_08_08_12_12_08_08_12_12_08:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,8,9,0,1,0,1,8,9,8,9,0,1,16,17,24,25,24,25,16,17,16,17,24,25,24,25,16,17]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,8,9,0,1,0,1,8,9,8,9,2,3,16,17,24,25,24,25,16,17,16,17,24,25,24,25,18,19]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_04_04_00_00_04_04_08_08_12_12_08_08_12_12_08:

 ;
 ; AVX2-LABEL: shuffle_v16i16_04_00_00_04_04_00_00_12_12_08_08_12_12_08_08_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,0,1,0,1,8,9,8,9,0,1,0,1,8,9,24,25,16,17,16,17,24,25,24,25,16,17,16,17,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[8,9,0,1,0,1,8,9,8,9,0,1,0,1,2,3,24,25,16,17,16,17,24,25,24,25,16,17,16,17,18,19]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_00_00_04_04_00_00_12_12_08_08_12_12_08_08_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_02_06_04_00_05_01_07_11_10_14_12_08_13_09_15_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,12,13,8,9,0,1,10,11,2,3,14,15,6,7,20,21,28,29,24,25,16,17,26,27,18,19,30,31,22,23]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_02_00_06_04_05_01_07_11_10_08_14_12_13_09_15_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,0,1,12,13,8,9,10,11,2,3,14,15,6,7,20,21,16,17,28,29,24,25,26,27,18,19,30,31,22,23]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_02_06_04_00_01_03_07_13_10_14_12_08_09_11_15_13:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7,8,9,10,11,12],ymm1[13],ymm0[14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[4,5,12,13,8,9,0,1,2,3,6,7,14,15,10,11,20,21,28,29,24,25,16,17,18,19,22,23,30,31,26,27]
 ; AVX2-NEXT:    retq
 ;

 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-SLOW-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
-; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7,28,29,28,29,30,31,26,27,18,19,28,29,24,25,22,23]
-; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
-; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,2,3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
-; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7,28,29,28,29,30,31,26,27,18,19,28,29,24,25,22,23]
-; AVX2-FAST-NEXT:    retq
+; AVX2-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,5,2,3,4,5,6,7]
+; AVX2-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,12,13,14,15,10,11,2,3,12,13,8,9,6,7,28,29,28,29,30,31,26,27,18,19,28,29,24,25,22,23]
+; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_06_06_07_05_01_06_04_11_14_14_15_13_09_14_12_11:
 ; AVX512VL:       # %bb.0:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_00_04_04_04_04_04_12_08_08_12_12_12_12_12_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,0,1,8,9,8,9,8,9,8,9,8,9,8,9,16,17,16,17,24,25,24,25,24,25,24,25,24,25,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,0,1,8,9,8,9,8,9,8,9,8,9,14,15,16,17,16,17,24,25,24,25,24,25,24,25,24,25,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_00_04_04_04_04_04_12_08_08_12_12_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_04_04_00_00_04_04_04_12_12_12_08_08_12_12_12_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,8,9,0,1,0,1,8,9,8,9,8,9,8,9,24,25,24,25,16,17,16,17,24,25,24,25,24,25,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[8,9,8,9,0,1,0,1,8,9,8,9,8,9,14,15,24,25,24,25,16,17,16,17,24,25,24,25,24,25,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_04_00_00_04_04_04_12_12_12_08_08_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_04_04_00_04_04_04_12_08_12_12_08_12_12_12_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,8,9,0,1,8,9,8,9,8,9,8,9,16,17,24,25,24,25,16,17,24,25,24,25,24,25,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,8,9,0,1,8,9,8,9,8,9,14,15,16,17,24,25,24,25,16,17,24,25,24,25,24,25,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_04_04_00_04_04_04_12_08_12_12_08_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_04_04_00_00_00_00_08_08_12_12_08_08_08_08_08:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpbroadcastw %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,8,9,0,1,0,1,0,1,0,1,0,1,16,17,24,25,24,25,16,17,16,17,16,17,16,17,16,17]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,8,9,0,1,0,1,0,1,0,1,14,15,16,17,24,25,24,25,16,17,16,17,16,17,16,17,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpslldq {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[16,17]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_04_04_00_00_00_00_08_08_12_12_08_08_08_08_08:

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,2,0,4,5,6,7,8,10,10,8,12,13,14,15]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,8,9,0,1,8,9,10,11,12,13,14,15,16,17,24,25,24,25,16,17,24,25,26,27,28,29,30,31]
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,8,9,0,1,8,9,10,11,12,13,14,15,16,17,24,25,24,25,16,17,24,25,26,27,28,29,30,31]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_04_04_00_04_05_06_15_08_12_12_08_12_13_14_15:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_uu_04_04_04_04_04_12_08_uu_12_12_12_12_12_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,8,9,8,9,8,9,8,9,8,9,8,9,16,17,18,19,24,25,24,25,24,25,24,25,24,25,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,u,u,255,255,255,255,255,255,255,255,255,255,0,0,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255>
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,2,3,8,9,8,9,8,9,8,9,8,9,14,15,16,17,18,19,24,25,24,25,24,25,24,25,24,25,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_uu_04_04_04_04_04_12_08_uu_12_12_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_04_04_uu_00_04_04_04_12_12_12_uu_08_12_12_12_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,8,9,8,9,0,1,8,9,8,9,8,9,8,9,24,25,24,25,24,25,16,17,24,25,24,25,24,25,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,255,255,u,u,255,255,255,255,255,255,255,255,0,0,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255>
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[8,9,8,9,8,9,0,1,8,9,8,9,8,9,14,15,24,25,24,25,24,25,16,17,24,25,24,25,24,25,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_04_uu_00_04_04_04_12_12_12_uu_08_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_uu_04_04_00_04_04_04_12_uu_12_12_08_12_12_12_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,8,9,0,1,8,9,8,9,8,9,8,9,16,17,24,25,24,25,16,17,24,25,24,25,24,25,24,25]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,255,255,255,255,255,255,255,255,255,255,255,255,0,0,u,u,255,255,255,255,255,255,255,255,255,255,255,255,255,255>
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,8,9,0,1,8,9,8,9,8,9,14,15,16,17,24,25,24,25,16,17,24,25,24,25,24,25,30,31]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpsllq $48, %ymm0, %ymm0
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_uu_04_04_00_04_04_04_12_uu_12_12_08_12_12_12_12:

 ;
 ; AVX2-LABEL: shuffle_v16i16_uu_uu_uu_uu_04_05_06_11_uu_uu_uu_uu_12_13_14_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,1,2,3]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[12,13,6,7,4,5,6,7,8,9,10,11,12,13,6,7,28,29,22,23,20,21,22,23,24,25,26,27,28,29,22,23]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,2,2,2]
+; 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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_uu_uu_uu_uu_04_05_06_11_uu_uu_uu_uu_12_13_14_11:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,8,9,10,11,12,13,6,7,16,17,18,19,20,21,30,31,24,25,26,27,28,29,22,23]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_04_05_06_03_00_01_02_15_12_13_14_11_08_09_10_15:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[8,9,10,11,12,13,6,7,0,1,2,3,4,5,14,15,24,25,26,27,28,29,22,23,16,17,18,19,20,21,30,31]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,6,7,0,1,2,3,4,5,2,3,24,25,26,27,28,29,22,23,16,17,18,19,20,21,18,19]
+; 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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_05_06_03_00_01_02_15_12_13_14_11_08_09_10_15:

 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-SLOW-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
-; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11,22,23,30,31,18,19,16,17,20,21,30,31,22,23,26,27]
-; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
-; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,1,6,3,4,5,6,7]
-; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm1, %ymm0
-; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11,22,23,30,31,18,19,16,17,20,21,30,31,22,23,26,27]
-; AVX2-FAST-NEXT:    retq
+; AVX2-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [0,1,6,3,4,5,6,7]
+; AVX2-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,14,15,2,3,0,1,4,5,14,15,6,7,10,11,22,23,30,31,18,19,16,17,20,21,30,31,22,23,26,27]
+; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_03_07_01_00_02_07_03_13_11_15_09_08_10_15_11_13:
 ; AVX512VL:       # %bb.0:

 ;
 ; AVX2-LABEL: shuffle_v16i16_00_16_01_17_02_18_03_27_08_24_09_25_10_26_11_27:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,255,255,0,0,u,u,u,u,u,u,u,u,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7,8,9,10],ymm2[11],ymm1[12,13,14,15]
 ; AVX2-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_00_20_01_21_02_22_03_31_08_28_09_29_10_30_11_31:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,8,9,2,3,10,11,4,5,12,13,6,7,14,15,16,17,24,25,18,19,26,27,20,21,28,29,22,23,30,31]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,2,3,10,11,4,5,12,13,6,7,14,15,16,17,24,25,18,19,26,27,20,21,28,29,22,23,30,31]
+; 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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_20_01_21_02_22_03_31_08_28_09_29_10_30_11_31:

 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm0[4,4,5,5,6,6,7,7,12,12,13,13,14,14,15,15]
 ; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm2 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6,7]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
 ; 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
 ;

 ; AVX2-LABEL: shuffle_v16i16_04_16_05_17_06_18_07_27_12_24_13_25_14_26_15_27:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[8,9,0,1,10,11,2,3,12,13,4,5,14,15,6,7,24,25,16,17,26,27,18,19,28,29,20,21,30,31,22,23]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[0,0,2,1,4,5,6,7,8,8,10,9,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,6,6,7,8,9,10,11,12,14,14,15]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6],ymm2[7]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]
 ; AVX2-SLOW-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-FAST-LABEL: shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[0,1,0,1,4,5,2,3,8,9,12,13,12,13,14,15,16,17,16,17,20,21,18,19,24,25,28,29,28,29,30,31]
-; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6],ymm2[7]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,2,3,6,7,12,13,10,11,14,15,14,15,16,17,18,19,18,19,22,23,28,29,26,27,30,31,30,31]
 ; AVX2-FAST-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-FAST-NEXT:    retq

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_25:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u>
-; AVX2-SLOW-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7,8],ymm2[9],ymm1[10,11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[8,9,8,9,4,5,10,11,0,1,0,1,12,13,2,3,24,25,24,25,20,21,26,27,16,17,16,17,28,29,18,19]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,6,7,8,9,9,11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,6,5,7,7,8,9,10,11,14,13,15,15]

 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_00_20_01_21_06_16_07_25_08_28_09_29_14_24_15_25:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u,255,255,255,255,u,u,u,u>
-; AVX2-FAST-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7,8],ymm2[9],ymm1[10,11,12,13,14,15]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[8,9,8,9,4,5,10,11,0,1,0,1,12,13,2,3,24,25,24,25,20,21,26,27,16,17,16,17,28,29,18,19]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,2,3,6,7,12,13,10,11,14,15,14,15,16,17,18,19,18,19,22,23,28,29,26,27,30,31,30,31]
 ; AVX2-FAST-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-SLOW-LABEL: shuffle_v16i16_01_00_17_16_03_02_19_26_09_08_25_24_11_10_27_26:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,0,0,255,255,u,u,u,u,u,u,u,u,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u>
-; AVX2-SLOW-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6,7,8,9],ymm2[10],ymm1[11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm1 = ymm1[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]

 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_01_00_17_16_03_02_19_26_09_08_25_24_11_10_27_26:
 ; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,0,0,255,255,u,u,u,u,u,u,u,u,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT:    vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,2,3]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4,5,6,7,8,9],ymm2[10],ymm1[11,12,13,14,15]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm1[2,3,0,1,2,3,0,1,8,9,10,11,6,7,4,5,18,19,16,17,18,19,16,17,24,25,26,27,22,23,20,21]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5,6,7,4,5,12,13,14,15,18,19,16,17,22,23,20,21,22,23,20,21,28,29,30,31]
 ; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]

 ;
 ; AVX2-LABEL: shuffle_v16i16_16_00_17_01_18_02_19_11_24_08_25_09_26_10_27_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = <255,255,255,255,255,255,0,0,u,u,u,u,u,u,u,u,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpblendvb %ymm3, %ymm0, %ymm2, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6,7,8,9,10],ymm2[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
 ; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm2 = ymm0[4,4,5,5,6,6,7,7,12,12,13,13,14,14,15,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6,7]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6],ymm0[7]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3],ymm1[4],ymm0[5],ymm1[6],ymm0[7],ymm1[8],ymm0[9],ymm1[10],ymm0[11],ymm1[12],ymm0[13],ymm1[14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-SLOW-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,1,3,4,5,6,7,8,10,9,11,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,6,5,7,8,9,10,11,12,14,13,15]
-; AVX2-SLOW-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-SLOW-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm1 = ymm0[0,1,4,5,2,3,6,7,8,9,12,13,10,11,14,15,16,17,20,21,18,19,22,23,24,25,28,29,26,27,30,31]
-; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-FAST-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,2,3,6,7,8,9,12,13,10,11,14,15,16,17,20,21,18,19,22,23,24,25,28,29,26,27,30,31]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:

 ;
 ; AVX2-LABEL: shuffle_v16i16_uu_uu_uu_uu_20_21_22_11_uu_uu_uu_uu_28_29_30_11:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,2]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7],ymm1[8,9,10,11,12,13,14],ymm0[15]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_00_01_02_21_20_21_22_11_08_09_10_29_28_29_30_11:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <255,255,255,255,255,255,0,0,255,255,255,255,255,255,u,u,255,255,255,255,255,255,255,255,255,255,255,255,255,255,u,u>
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,10,11,8,9,10,11,12,13,6,7,16,17,18,19,20,21,26,27,24,25,26,27,28,29,22,23]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_25:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm1, %xmm1
-; AVX2-SLOW-NEXT:    vpbroadcastd %xmm1, %ymm1
+; AVX2-SLOW-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
 ; AVX2-SLOW-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,1,2,1,4,5,6,7,8,9,10,9,12,13,14,15]
 ; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,7,7,8,9,10,11,12,13,15,15]
 ; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]

 ;
 ; AVX2-FAST-LABEL: shuffle_v16i16_uu_uu_uu_01_uu_05_07_25_uu_uu_uu_09_uu_13_15_25:
 ; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [4,5,6,4,4,5,6,4]
+; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,2,3,8,9,10,11,14,15,14,15,16,17,18,19,20,21,18,19,24,25,26,27,30,31,30,31]
-; AVX2-FAST-NEXT:    vextracti128 $1, %ymm1, %xmm1
-; AVX2-FAST-NEXT:    vpbroadcastd %xmm1, %ymm1
 ; AVX2-FAST-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; AVX2-FAST-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_21_22_23_00_01_02_03_12_29_30_31_08_09_10_11_12:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7,8,9,10,11],ymm1[12],ymm0[13,14,15]
 ; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_05_06_07_00_01_02_03_12_13_14_15_08_09_10_11_12:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7,8,9,10,11],ymm1[12],ymm0[13,14,15]
 ; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_19_20_21_22_23_00_01_10_27_28_29_30_31_08_09_10:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7],ymm0[8,9,10],ymm1[11,12,13,14,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; 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:    vpalignr {{.*#+}} ymm0 = ymm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20,21]
 ; AVX2-NEXT:    retq
 ;

 ;
 ; AVX2-LABEL: shuffle_v16i16_03_04_05_06_07_00_01_10_11_12_13_14_15_08_09_10:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; 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:    vpalignr {{.*#+}} ymm0 = ymm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20,21]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_03_04_05_06_07_16_17_26_11_12_13_14_15_24_25_26:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; 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:    vpalignr {{.*#+}} ymm0 = ymm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20,21]
 ; AVX2-NEXT:    retq
 ;

 ; AVX2-LABEL: shuffle_v16i16_05_06_07_16_17_18_19_28_13_14_15_24_25_26_27_28:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,2,3]
+; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7,8,9,10,11],ymm1[12],ymm0[13,14,15]
 ; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[10,11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,26,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25]
 ; AVX2-NEXT:    retq
 ;

test/CodeGen/X86/vector-shuffle-256-v32.ll

 ; AVX2-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpbroadcastb %xmm0, %xmm0
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpslldq {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16]
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0]
+; AVX2-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:
 ; AVX512VLBW:       # %bb.0:
 ; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX512VLBW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
 ; AVX512VLBW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX512VLBW-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX512VLBW-NEXT:    vpbroadcastb %xmm0, %xmm0
-; AVX512VLBW-NEXT:    movl $32767, %eax # imm = 0x7FFF
+; AVX512VLBW-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
+; AVX512VLBW-NEXT:    vpslldq {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm1[16]
+; AVX512VLBW-NEXT:    movl $-2147450880, %eax # imm = 0x80008000
 ; AVX512VLBW-NEXT:    kmovd %eax, %k1
-; AVX512VLBW-NEXT:    vmovdqu8 %ymm0, %ymm1 {%k1}
-; AVX512VLBW-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512VLBW-NEXT:    vmovdqu8 %ymm1, %ymm0 {%k1}
 ; AVX512VLBW-NEXT:    retq
 ;
 ; AVX512VLVBMI-LABEL: shuffle_v32i8_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_16_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00_00:

 ;
 ; AVX2-LABEL: shuffle_v32i8_31_32_33_34_35_36_37_38_39_40_41_42_43_44_45_46_47_48_49_50_51_52_53_54_55_56_57_58_59_60_61_62:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0]
-; AVX2-NEXT:    vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm1[15],ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14],ymm1[31],ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28,29,30]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
+; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14],ymm0[31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27,28,29,30]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VLBW-LABEL: shuffle_v32i8_31_32_33_34_35_36_37_38_39_40_41_42_43_44_45_46_47_48_49_50_51_52_53_54_55_56_57_58_59_60_61_62:
 ; AVX512VLBW:       # %bb.0:
-; AVX512VLBW-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; AVX512VLBW-NEXT:    kmovd %eax, %k1
-; AVX512VLBW-NEXT:    vmovdqu8 %ymm0, %ymm1 {%k1}
-; AVX512VLBW-NEXT:    vpermq {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX512VLBW-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
 ; AVX512VLBW-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14],ymm0[31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27,28,29,30]
 ; AVX512VLBW-NEXT:    retq
 ;

test/CodeGen/X86/vector-shuffle-256-v4.ll

 define <4 x double> @shuffle_v4f64_0462(<4 x double> %a, <4 x double> %b) {
 ; ALL-LABEL: shuffle_v4f64_0462:
 ; ALL:       # %bb.0:
-; ALL-NEXT:    vmovddup {{.*#+}} ymm1 = ymm1[0,0,2,2]
-; ALL-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
-; ALL-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3]
+; ALL-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; ALL-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; ALL-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
 ; ALL-NEXT:    retq
   %shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 4, i32 6, i32 2>
   ret <4 x double> %shuffle

 define <4 x i64> @shuffle_v4i64_1z3z(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_1z3z:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vshufpd {{.*#+}} xmm1 = xmm1[1],xmm0[0]
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_1z3z:

 define <4 x i64> @shuffle_v4i64_1234(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: shuffle_v4i64_1234:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1]
 ; AVX1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[3],ymm1[2]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v4i64_1234:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,2,3,0]
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1]
+; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm0[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i64_1234:

 define <4 x double> @add_v4f64_0246_1357(<4 x double> %a, <4 x double> %b) {
 ; AVX1-LABEL: add_v4f64_0246_1357:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm2[0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
-; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm0[0],xmm4[0]
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3]
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-NEXT:    vaddpd %ymm0, %ymm3, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
+; AVX1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v4f64_0246_1357:

 define <4 x double> @add_v4f64_4602_5713(<4 x double> %a, <4 x double> %b) {
 ; AVX1-LABEL: add_v4f64_4602_5713:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm0[0],xmm2[0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
-; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm1[0],xmm4[0]
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3]
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
-; AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX1-NEXT:    vaddpd %ymm0, %ymm3, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
+; AVX1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v4f64_4602_5713:

 define <4 x i64> @add_v4i64_0246_1357(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: add_v4i64_0246_1357:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm2[0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
-; AVX1-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm0[0],xmm4[0]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm2
-; AVX1-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpaddq %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v4i64_0246_1357:

 define <4 x i64> @add_v4i64_4602_5713(<4 x i64> %a, <4 x i64> %b) {
 ; AVX1-LABEL: add_v4i64_4602_5713:
 ; AVX1:       # %bb.0: # %entry
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX1-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm0[0],xmm2[0]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
-; AVX1-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm1[0],xmm4[0]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm2
-; AVX1-NEXT:    vpaddq %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpaddq %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v4i64_4602_5713:

test/CodeGen/X86/vector-shuffle-256-v8.ll

 define <8 x float> @shuffle_v8f32_00040000(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_00040000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,3]
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[0,0],ymm1[4,4],ymm0[4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[2,0],ymm0[4,4],ymm1[6,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8f32_00040000:

 define <8 x float> @shuffle_v8f32_c348cda0(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_c348cda0:
 ; AVX1:       # %bb.0:
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,0],ymm1[4,5],ymm3[6,4]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[0,0],ymm0[4,7],ymm2[4,4]
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5],ymm2[6,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3,4,5,6],ymm0[7]
 ; AVX1-NEXT:    retq
 ;

 define <8 x float> @PR21138(<8 x float> %truc, <8 x float> %tchose) {
 ; AVX1-LABEL: PR21138:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm2[1,3],ymm0[5,7],ymm2[5,7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: PR21138:

 define <8 x i32> @shuffle_v8i32_00040000(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_00040000:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,0,0,3]
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,0,0,0,4,4,4,4]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[0,0],ymm1[4,4],ymm0[4,4]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[2,0],ymm0[4,4],ymm1[6,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_00040000:

 define <8 x i32> @shuffle_v8i32_12345678(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_12345678:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7]
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm0[3,0],ymm1[4,4],ymm0[7,4]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,2],ymm1[2,0],ymm0[5,6],ymm1[6,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i32_12345678:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7]
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm1 = [1,2,3,4,5,6,7,0]
-; AVX2-NEXT:    vpermps %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1]
+; AVX2-NEXT:    vpalignr {{.*#+}} ymm0 = ymm0[4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0,1,2,3],ymm0[20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16,17,18,19]
 ; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v8i32_12345678:

 define <8 x float> @add_v8f32_02468ACE_13579BDF(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: add_v8f32_02468ACE_13579BDF:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[0,2],xmm2[0,2]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
-; AVX1-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[0,2],xmm4[0,2]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm4[1,3]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-NEXT:    vaddps %ymm0, %ymm3, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[0,2],ymm2[0,2],ymm0[4,6],ymm2[4,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm2[1,3],ymm0[5,7],ymm2[5,7]
+; AVX1-NEXT:    vaddps %ymm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v8f32_02468ACE_13579BDF:

 define <8 x float> @add_v8f32_8ACE0246_9BDF1357(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: add_v8f32_8ACE0246_9BDF1357:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[0,2],xmm2[0,2]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
-; AVX1-NEXT:    vshufps {{.*#+}} xmm5 = xmm1[0,2],xmm4[0,2]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm4[1,3]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-NEXT:    vaddps %ymm0, %ymm3, %ymm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[0,2],ymm2[0,2],ymm0[4,6],ymm2[4,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm2[1,3],ymm0[5,7],ymm2[5,7]
+; AVX1-NEXT:    vaddps %ymm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v8f32_8ACE0246_9BDF1357:

 define <8 x i32> @add_v8i32_02468ACE_13579BDF(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: add_v8i32_02468ACE_13579BDF:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[0,2],xmm2[0,2]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
-; AVX1-NEXT:    vshufps {{.*#+}} xmm5 = xmm0[0,2],xmm4[0,2]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm4[1,3]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm2
-; AVX1-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpaddd %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[0,2],ymm2[0,2],ymm0[4,6],ymm2[4,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm2[1,3],ymm0[5,7],ymm2[5,7]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpaddd %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v8i32_02468ACE_13579BDF:

 define <8 x i32> @add_v8i32_8ACE0246_9BDF1357(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: add_v8i32_8ACE0246_9BDF1357:
 ; AVX1:       # %bb.0: # %entry
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vshufps {{.*#+}} ymm1 = ymm0[0,2],ymm2[0,2],ymm0[4,6],ymm2[4,6]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,3],ymm2[1,3],ymm0[5,7],ymm2[5,7]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[0,2],xmm2[0,2]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
-; AVX1-NEXT:    vshufps {{.*#+}} xmm5 = xmm1[0,2],xmm4[0,2]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,3],xmm4[1,3]
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm2
-; AVX1-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpaddd %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpaddd %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: add_v8i32_8ACE0246_9BDF1357:

test/CodeGen/X86/vector-shuffle-512-v64.ll

 define <64 x i8> @shuffle_v64i8_63_zz_61_zz_59_zz_57_zz_55_zz_53_zz_51_zz_49_zz_47_zz_45_zz_43_zz_41_zz_39_zz_37_zz_35_zz_33_zz_31_zz_29_zz_27_zz_25_zz_23_zz_21_zz_19_zz_17_zz_15_zz_13_zz_11_zz_9_zz_7_zz_5_zz_3_zz_1_zz(<64 x i8> %a) {
 ; AVX512F-LABEL: shuffle_v64i8_63_zz_61_zz_59_zz_57_zz_55_zz_53_zz_51_zz_49_zz_47_zz_45_zz_43_zz_41_zz_39_zz_37_zz_35_zz_33_zz_31_zz_29_zz_27_zz_25_zz_23_zz_21_zz_19_zz_17_zz_15_zz_13_zz_11_zz_9_zz_7_zz_5_zz_3_zz_1_zz:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = <15,u,13,u,11,u,9,u,7,u,5,u,3,u,1,u,15,u,13,u,11,u,9,u,7,u,5,u,3,u,1,u>
-; AVX512F-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
 ; AVX512F-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [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]
-; AVX512F-NEXT:    vpand %ymm4, %ymm1, %ymm2
-; AVX512F-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [15,128,13,128,11,128,9,128,7,128,5,128,3,128,1,128,15,128,13,128,11,128,9,128,7,128,5,128,3,128,1,128]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm1, %ymm2
 ; AVX512F-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX512F-NEXT:    vpand %ymm4, %ymm0, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm0, %ymm1
 ; AVX512F-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512BW-LABEL: shuffle_v64i8_63_zz_61_zz_59_zz_57_zz_55_zz_53_zz_51_zz_49_zz_47_zz_45_zz_43_zz_41_zz_39_zz_37_zz_35_zz_33_zz_31_zz_29_zz_27_zz_25_zz_23_zz_21_zz_19_zz_17_zz_15_zz_13_zz_11_zz_9_zz_7_zz_5_zz_3_zz_1_zz:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm1 = <15,u,13,u,11,u,9,u,7,u,5,u,3,u,1,u,15,u,13,u,11,u,9,u,7,u,5,u,3,u,1,u>
-; AVX512BW-NEXT:    vpshufb %ymm1, %ymm0, %ymm2
-; AVX512BW-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = [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]
-; AVX512BW-NEXT:    vpand %ymm3, %ymm2, %ymm2
+; AVX512BW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,128,13,128,11,128,9,128,7,128,5,128,3,128,1,128,15,128,13,128,11,128,9,128,7,128,5,128,3,128,1,128]
+; AVX512BW-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
-; AVX512BW-NEXT:    vpshufb %ymm1, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX512BW-NEXT:    vpand %ymm3, %ymm0, %ymm0
-; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
+; AVX512BW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: shuffle_v64i8_63_zz_61_zz_59_zz_57_zz_55_zz_53_zz_51_zz_49_zz_47_zz_45_zz_43_zz_41_zz_39_zz_37_zz_35_zz_33_zz_31_zz_29_zz_27_zz_25_zz_23_zz_21_zz_19_zz_17_zz_15_zz_13_zz_11_zz_9_zz_7_zz_5_zz_3_zz_1_zz:
 ; AVX512DQ:       # %bb.0:
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm3 = <15,u,13,u,11,u,9,u,7,u,5,u,3,u,1,u,15,u,13,u,11,u,9,u,7,u,5,u,3,u,1,u>
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm1
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm4 = [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]
-; AVX512DQ-NEXT:    vpand %ymm4, %ymm1, %ymm2
-; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
+; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [15,128,13,128,11,128,9,128,7,128,5,128,3,128,1,128,15,128,13,128,11,128,9,128,7,128,5,128,3,128,1,128]
+; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm1, %ymm2
 ; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpand %ymm4, %ymm0, %ymm1
+; AVX512DQ-NEXT:    vpshufb %ymm3, %ymm0, %ymm1
 ; AVX512DQ-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512DQ-NEXT:    retq
 ;

 define <64 x i8> @shuffle_v64i8_63_64_61_66_59_68_57_70_55_72_53_74_51_76_49_78_47_80_45_82_43_84_41_86_39_88_37_90_35_92_33_94_31_96_29_98_27_100_25_102_23_104_21_106_19_108_17_110_15_112_13_114_11_116_9_118_7_120_5_122_3_124_1_126(<64 x i8> %a, <64 x i8> %b) {
 ; AVX512F-LABEL: shuffle_v64i8_63_64_61_66_59_68_57_70_55_72_53_74_51_76_49_78_47_80_45_82_43_84_41_86_39_88_37_90_35_92_33_94_31_96_29_98_27_100_25_102_23_104_21_106_19_108_17_110_15_112_13_114_11_116_9_118_7_120_5_122_3_124_1_126:
 ; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
 ; AVX512F-NEXT:    vpbroadcastw {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
 ; AVX512F-NEXT:    vpblendvb %ymm4, %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512F-NEXT:    vpblendvb %ymm4, %ymm1, %ymm2, %ymm1
 ; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm5 = [15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14,15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14]
 ; AVX512F-NEXT:    vpshufb %ymm5, %ymm1, %ymm2
+; AVX512F-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX512F-NEXT:    vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX512F-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm0
 ; AVX512F-NEXT:    vpshufb %ymm5, %ymm0, %ymm1
 ; AVX512F-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512F-NEXT:    retq

 ; AVX512BW-LABEL: shuffle_v64i8_63_64_61_66_59_68_57_70_55_72_53_74_51_76_49_78_47_80_45_82_43_84_41_86_39_88_37_90_35_92_33_94_31_96_29_98_27_100_25_102_23_104_21_106_19_108_17_110_15_112_13_114_11_116_9_118_7_120_5_122_3_124_1_126:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vextracti64x4 $1, %zmm1, %ymm2
-; AVX512BW-NEXT:    vpbroadcastw {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512BW-NEXT:    vpblendvb %ymm3, %ymm2, %ymm0, %ymm2
-; AVX512BW-NEXT:    vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX512BW-NEXT:    vpblendvb %ymm3, %ymm2, %ymm4, %ymm2
-; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm4 = [15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14,15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14]
-; AVX512BW-NEXT:    vpshufb %ymm4, %ymm2, %ymm2
+; AVX512BW-NEXT:    vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX512BW-NEXT:    vpbroadcastw {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512BW-NEXT:    vpblendvb %ymm4, %ymm2, %ymm3, %ymm2
+; AVX512BW-NEXT:    vmovdqa {{.*#+}} ymm3 = [15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14,15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14]
+; AVX512BW-NEXT:    vpshufb %ymm3, %ymm2, %ymm2
 ; AVX512BW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
-; AVX512BW-NEXT:    vpblendvb %ymm3, %ymm1, %ymm0, %ymm0
-; AVX512BW-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX512BW-NEXT:    vpblendvb %ymm3, %ymm0, %ymm1, %ymm0
-; AVX512BW-NEXT:    vpshufb %ymm4, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX512BW-NEXT:    vpblendvb %ymm4, %ymm1, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpshufb %ymm3, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: shuffle_v64i8_63_64_61_66_59_68_57_70_55_72_53_74_51_76_49_78_47_80_45_82_43_84_41_86_39_88_37_90_35_92_33_94_31_96_29_98_27_100_25_102_23_104_21_106_19_108_17_110_15_112_13_114_11_116_9_118_7_120_5_122_3_124_1_126:
 ; AVX512DQ:       # %bb.0:
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
 ; AVX512DQ-NEXT:    vpbroadcastw {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
 ; AVX512DQ-NEXT:    vpblendvb %ymm4, %ymm2, %ymm1, %ymm1
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendvb %ymm4, %ymm1, %ymm2, %ymm1
 ; AVX512DQ-NEXT:    vmovdqa {{.*#+}} ymm5 = [15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14,15,0,13,2,11,4,9,6,7,8,5,10,3,12,1,14]
 ; AVX512DQ-NEXT:    vpshufb %ymm5, %ymm1, %ymm2
+; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
 ; AVX512DQ-NEXT:    vpblendvb %ymm4, %ymm3, %ymm0, %ymm0
-; AVX512DQ-NEXT:    vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX512DQ-NEXT:    vpblendvb %ymm4, %ymm0, %ymm1, %ymm0
 ; AVX512DQ-NEXT:    vpshufb %ymm5, %ymm0, %ymm1
 ; AVX512DQ-NEXT:    vmovdqa %ymm2, %ymm0
 ; AVX512DQ-NEXT:    retq

test/CodeGen/X86/vector-shuffle-avx512.ll

 ; SKX64-LABEL: expand2:
 ; SKX64:       # %bb.0:
 ; SKX64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; SKX64-NEXT:    movb $9, %al
-; SKX64-NEXT:    kmovd %eax, %k1
-; SKX64-NEXT:    vexpandpd %ymm0, %ymm0 {%k1} {z}
+; SKX64-NEXT:    vperm2f128 {{.*#+}} ymm1 = zero,zero,ymm0[0,1]
+; SKX64-NEXT:    vmovaps %xmm0, %xmm0
+; SKX64-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; SKX64-NEXT:    retq
 ;
 ; KNL64-LABEL: expand2:
 ; KNL64:       # %bb.0:
 ; KNL64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL64-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
-; KNL64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; KNL64-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5],ymm0[6,7]
+; KNL64-NEXT:    vperm2f128 {{.*#+}} ymm1 = zero,zero,ymm0[0,1]
+; KNL64-NEXT:    vmovaps %xmm0, %xmm0
+; KNL64-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; KNL64-NEXT:    retq
 ;
 ; SKX32-LABEL: expand2:
 ; SKX32:       # %bb.0:
 ; SKX32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; SKX32-NEXT:    movb $9, %al
-; SKX32-NEXT:    kmovd %eax, %k1
-; SKX32-NEXT:    vexpandpd %ymm0, %ymm0 {%k1} {z}
+; SKX32-NEXT:    vperm2f128 {{.*#+}} ymm1 = zero,zero,ymm0[0,1]
+; SKX32-NEXT:    vmovaps %xmm0, %xmm0
+; SKX32-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; SKX32-NEXT:    retl
 ;
 ; KNL32-LABEL: expand2:
 ; KNL32:       # %bb.0:
 ; KNL32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL32-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
-; KNL32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; KNL32-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5],ymm0[6,7]
+; KNL32-NEXT:    vperm2f128 {{.*#+}} ymm1 = zero,zero,ymm0[0,1]
+; KNL32-NEXT:    vmovaps %xmm0, %xmm0
+; KNL32-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; KNL32-NEXT:    retl
    %res = shufflevector <2 x double> %a, <2 x double> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 2, i32 1>
    ret <4 x double> %res

 ; KNL64-LABEL: expand4:
 ; KNL64:       # %bb.0:
 ; KNL64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL64-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
-; KNL64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; KNL64-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5],ymm0[6,7]
+; KNL64-NEXT:    vperm2f128 {{.*#+}} ymm1 = zero,zero,ymm0[0,1]
+; KNL64-NEXT:    vmovaps %xmm0, %xmm0
+; KNL64-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; KNL64-NEXT:    retq
 ;
 ; SKX32-LABEL: expand4:

 ; KNL32-LABEL: expand4:
 ; KNL32:       # %bb.0:
 ; KNL32-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
-; KNL32-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,1,2,1]
-; KNL32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; KNL32-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5],ymm0[6,7]
+; KNL32-NEXT:    vperm2f128 {{.*#+}} ymm1 = zero,zero,ymm0[0,1]
+; KNL32-NEXT:    vmovaps %xmm0, %xmm0
+; KNL32-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; KNL32-NEXT:    retl
    %res = shufflevector <2 x i64> zeroinitializer, <2 x i64> %a, <4 x i32> <i32 2, i32 0, i32 0, i32 3>
    ret <4 x i64> %res

test/CodeGen/X86/vector-shuffle-combining.ll

 ;
 ; AVX1-LABEL: PR22412:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[3,2],ymm0[5,4],ymm1[7,6]
+; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm2[3,2],ymm0[5,4],ymm2[7,6]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-SLOW-LABEL: PR22412:
-; AVX2-SLOW:       # %bb.0: # %entry
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[1,0,3,2,5,4,7,6]
-; AVX2-SLOW-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,1]
-; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: PR22412:
-; AVX2-FAST:       # %bb.0: # %entry
-; AVX2-FAST-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT:    vmovaps {{.*#+}} ymm1 = [1,0,7,6,5,4,3,2]
-; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm1, %ymm0
-; AVX2-FAST-NEXT:    retq
+; AVX2-LABEL: PR22412:
+; AVX2:       # %bb.0: # %entry
+; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[1,0],ymm1[3,2],ymm0[5,4],ymm1[7,6]
+; AVX2-NEXT:    retq
 entry:
   %s1 = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 1, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   %s2 = shufflevector <8 x float> %s1, <8 x float> undef, <8 x i32> <i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2>