[X86][SSE] SimplifyDemandedVectorEltsForTargetNode - add general shuffle combining support

[RKSimon]

Summary

This patch uses partial DemandedElts masks to further simplify target shuffle chains and finally starts making target shuffle combining part of SimplifyDemandedBits/SimplifyDemandedVectorElts.

We already manage this for Depth == 0 cases, where combineX86ShuffleChain would early-out if the shuffle combined to the same op, but the patch generalizes this by manipulating the depth handling of combineX86ShufflesRecursively - calling with a new Depth = 0 and reducing the maximum shuffle combine depth accordingly.

[xbolva00]

Can this patch solve bad codegen for 'f5'?

https://godbolt.org/z/3YpVg-

[craig.topper]

In D66004#1647773, @xbolva00 wrote:

Can this patch solve bad codegen for 'f5'?

https://godbolt.org/z/3YpVg-

I don't think any amount of shuffle combining is going to recover that. I think we need to look at lowerShuffleAsLanePermuteAndRepeatedMask

[craig.topper]

In D66004#1647822, @craig.topper wrote:

In D66004#1647773, @xbolva00 wrote:

Can this patch solve bad codegen for 'f5'?

https://godbolt.org/z/3YpVg-

I don't think any amount of shuffle combining is going to recover that. I think we need to look at lowerShuffleAsLanePermuteAndRepeatedMask

Nevermind, that won't fix it. The two lanes have different controls for the shufpd in gcc's code. I think we need a new strategy.

[xbolva00]

They improved f1 and f2 cases to save one instruction wrt LLVM’s codegen

https://gcc.gnu.org/ml/gcc-patches/2019-08/msg01952.html

[lebedev.ri]

Rebase this?

[RKSimon]

WIP - PR27854 and PR43024 need to be finished first.

[RKSimon]

rebase - still work to do to improve unpck vs insertps selection

[RKSimon]

rebase

[RKSimon]

rebase (still looking at this)

[RKSimon]

rebase

[xbolva00]

Status of this patch?

[RKSimon]

In D66004#2078352, @xbolva00 wrote:

Status of this patch?

I'm still looking at this - it both affects and is affected by so much code its a yak shaving nightmare to handle it all.

The big remaining issue is the loss of INSERTPS for some BUILDVECTOR patterns, which means we lose load folding on SSE41+ targets.

Additionally, we need to stop creating nodes on the fly inside combineX86ShufflesRecursively (see PR45974) as this screws up hasOneUse checks which are often vital in SimplifyDemandedBits/SimplifyDemandedVectorElts.

[RKSimon]

rebased - the variable shuffle mask simplification has helped but theres still some work to do - next is to fix the movhlps(x,movss) vs insertps regression

[RKSimon]

fix movhlps(x,scalar_to_vector(y)) -> insertps handling

[xbolva00]

llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll

5637 ↗

(On Diff #272316)

Regression

llvm/test/CodeGen/X86/buildvec-insertvec.ll

164

Extra instructio (vs sse2)

llvm/test/CodeGen/X86/masked_expandload.ll

1302

Regression

llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll

559

Worse

648

Ouch

[RKSimon]

cheers @xbolva00 for what its worth, this is still a WIP, and SimplifyDemanded improvements do have a tendency to expose existing limitations (as well as find some awesome improvements.....)

llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll

5637 ↗

(On Diff #272316)

The original vzext_movl (which isel would consume into the movd) has been replaced with a zero_extend_vector_inreg.

llvm/test/CodeGen/X86/buildvec-insertvec.ll

164

The original vzext_movl (which isel would consume into the movd) has been replaced with a zero_extend_vector_inreg.

llvm/test/CodeGen/X86/masked_expandload.ll

1302

We've managed to change the order of load combines and we prematurely end up with different element sizes which we then can't combine together later. This is an existing problem, we've just exposed it more.

llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll

559

still looking at these regressions - but we've always been very weak at shuffle combining different src/dst vector widths.

[RKSimon]

Fixed the vpmovzx regressions

[RKSimon]

rebase

[RKSimon]

rebase

[RKSimon]

rebase

[RKSimon]

rebase

[RKSimon]

rebase

[RKSimon]

All regressions should now be covered.

[xbolva00]

Amazing work Simon!

[RKSimon]

ping - I think this is ready for commital now, I delayed pinging as I knew I'd be away on holiday this week and a patch like this is likely to cause a few unexpected regressions that will need addressing.

[craig.topper]

LGTM

llvm/lib/Target/X86/X86ISelLowering.cpp

                               bool AllowFloatDomain, bool AllowIntDomain,
                               SDValue &V1, const SDLoc &DL, SelectionDAG &DAG,
                               const X86Subtarget &Subtarget, unsigned &Shuffle,
                               MVT &SrcVT, MVT &DstVT) {
   unsigned NumMaskElts = Mask.size();
   unsigned MaskEltSize = MaskVT.getScalarSizeInBits();
 
   // Match against a VZEXT_MOVL vXi32 zero-extending instruction.
-  if (MaskEltSize == 32 && isUndefOrEqual(Mask[0], 0) &&
-      isUndefOrZero(Mask[1]) && isUndefInRange(Mask, 2, NumMaskElts - 2)) {
+  if (MaskEltSize == 32 && Mask[0] == 0) {
+    if (isUndefOrZero(Mask[1]) && isUndefInRange(Mask, 2, NumMaskElts - 2)) {
       Shuffle = X86ISD::VZEXT_MOVL;
       SrcVT = DstVT = !Subtarget.hasSSE2() ? MVT::v4f32 : MaskVT;
       return true;
+    }
+    if (V1.getOpcode() == ISD::SCALAR_TO_VECTOR &&
+        isUndefOrZeroInRange(Mask, 1, NumMaskElts - 1)) {
+      Shuffle = X86ISD::VZEXT_MOVL;
+      SrcVT = DstVT = !Subtarget.hasSSE2() ? MVT::v4f32 : MaskVT;
+      return true;
+    }
   }
 
   // Match against a ANY/ZERO_EXTEND_VECTOR_INREG instruction.
   // TODO: Add 512-bit vector support (split AVX512F and AVX512BW).
   if (AllowIntDomain && ((MaskVT.is128BitVector() && Subtarget.hasSSE41()) ||
                          (MaskVT.is256BitVector() && Subtarget.hasInt256()))) {
     unsigned MaxScale = 64 / MaskEltSize;
     for (unsigned Scale = 2; Scale <= MaxScale; Scale *= 2) {

       return DAG.getBitcast(RootVT, Res);
     }
   }
 
   // Attempt to combine to INSERTPS, but only if the inserted element has come
   // from a scalar.
   // TODO: Handle other insertions here as well?
   if (!UnaryShuffle && AllowFloatDomain && RootSizeInBits == 128 &&
-      MaskEltSizeInBits == 32 && Subtarget.hasSSE41() &&
-      !isTargetShuffleEquivalent(Mask, {4, 1, 2, 3})) {
+      Subtarget.hasSSE41() && !isTargetShuffleEquivalent(Mask, {4, 1, 2, 3})) {
+    if (MaskEltSizeInBits == 32) {
       SDValue SrcV1 = V1, SrcV2 = V2;
-    if (matchShuffleAsInsertPS(SrcV1, SrcV2, PermuteImm, Zeroable, Mask, DAG) &&
-        SrcV2.getOpcode() == ISD::SCALAR_TO_VECTOR) {
+      if (matchShuffleAsInsertPS(SrcV1, SrcV2, PermuteImm, Zeroable, Mask,
+                                 DAG) &&
+          SrcV2.getOpcode() == ISD::SCALAR_TO_VECTOR) {
+        if (Depth == 0 && Root.getOpcode() == X86ISD::INSERTPS)
+          return SDValue(); // Nothing to do!
+        Res = DAG.getNode(X86ISD::INSERTPS, DL, MVT::v4f32,
+                          DAG.getBitcast(MVT::v4f32, SrcV1),
+                          DAG.getBitcast(MVT::v4f32, SrcV2),
+                          DAG.getTargetConstant(PermuteImm, DL, MVT::i8));
+        return DAG.getBitcast(RootVT, Res);
+      }
+    }
+    if (MaskEltSizeInBits == 64 && isTargetShuffleEquivalent(Mask, {0, 2}) &&
+        V2.getOpcode() == ISD::SCALAR_TO_VECTOR &&
+        V2.getScalarValueSizeInBits() <= 32) {
       if (Depth == 0 && Root.getOpcode() == X86ISD::INSERTPS)
         return SDValue(); // Nothing to do!
+      PermuteImm = (/*DstIdx*/2 << 4) | (/*SrcIdx*/0 << 0);
       Res = DAG.getNode(X86ISD::INSERTPS, DL, MVT::v4f32,
-                        DAG.getBitcast(MVT::v4f32, SrcV1),
-                        DAG.getBitcast(MVT::v4f32, SrcV2),
+                        DAG.getBitcast(MVT::v4f32, V1),
+                        DAG.getBitcast(MVT::v4f32, V2),
                         DAG.getTargetConstant(PermuteImm, DL, MVT::i8));
       return DAG.getBitcast(RootVT, Res);
     }
   }
 
   SDValue NewV1 = V1; // Save operands in case early exit happens.
   SDValue NewV2 = V2;
   if (matchBinaryShuffle(MaskVT, Mask, AllowFloatDomain, AllowIntDomain, NewV1,

   MVT MaskVT = MVT::getVectorVT(MaskSVT, NumMaskElts);
   if (!DAG.getTargetLoweringInfo().isTypeLegal(MaskVT))
     return SDValue();
 
   SDValue CstOp = getConstVector(ConstantBitData, UndefElts, MaskVT, DAG, DL);
   return DAG.getBitcast(VT, CstOp);
 }
 
+namespace llvm {
+  namespace X86 {
+    enum {
+      MaxShuffleCombineDepth = 8
+    };
+  };
+}; // namespace llvm
+
 /// Fully generic combining of x86 shuffle instructions.
 ///
 /// This should be the last combine run over the x86 shuffle instructions. Once
 /// they have been fully optimized, this will recursively consider all chains
 /// of single-use shuffle instructions, build a generic model of the cumulative
 /// shuffle operation, and check for simpler instructions which implement this
 /// operation. We use this primarily for two purposes:
 ///

 ///
 /// FIXME: We will currently miss some cases where the redundant shuffling
 /// would simplify under the threshold for PSHUFB formation because of
 /// combine-ordering. To fix this, we should do the redundant instruction
 /// combining in this recursive walk.
 static SDValue combineX86ShufflesRecursively(
     ArrayRef<SDValue> SrcOps, int SrcOpIndex, SDValue Root,
     ArrayRef<int> RootMask, ArrayRef<const SDNode *> SrcNodes, unsigned Depth,
-    bool HasVariableMask, bool AllowVariableMask, SelectionDAG &DAG,
-    const X86Subtarget &Subtarget) {
+    unsigned MaxDepth, bool HasVariableMask, bool AllowVariableMask,
+    SelectionDAG &DAG, const X86Subtarget &Subtarget) {
   assert(RootMask.size() > 0 &&
          (RootMask.size() > 1 || (RootMask[0] == 0 && SrcOpIndex == 0)) &&
          "Illegal shuffle root mask");
   assert(Root.getSimpleValueType().isVector() &&
          "Shuffles operate on vector types!");
   unsigned RootSizeInBits = Root.getSimpleValueType().getSizeInBits();
 
   // Bound the depth of our recursive combine because this is ultimately
   // quadratic in nature.
-  const unsigned MaxRecursionDepth = 8;
-  if (Depth >= MaxRecursionDepth)
+  if (Depth >= MaxDepth)
     return SDValue();
 
   // Directly rip through bitcasts to find the underlying operand.
   SDValue Op = SrcOps[SrcOpIndex];
   Op = peekThroughOneUseBitcasts(Op);
 
   MVT VT = Op.getSimpleValueType();
   if (!VT.isVector())

 
   // See if we can recurse into each shuffle source op (if it's a target
   // shuffle). The source op should only be generally combined if it either has
   // a single use (i.e. current Op) or all its users have already been combined,
   // if not then we can still combine but should prevent generation of variable
   // shuffles to avoid constant pool bloat.
   // Don't recurse if we already have more source ops than we can combine in
   // the remaining recursion depth.
-  if (Ops.size() < (MaxRecursionDepth - Depth)) {
+  if (Ops.size() < (MaxDepth - Depth)) {
     for (int i = 0, e = Ops.size(); i < e; ++i) {
       // For empty roots, we need to resolve zeroable elements before combining
       // them with other shuffles.
       SmallVector<int, 64> ResolvedMask = Mask;
       if (EmptyRoot)
         resolveTargetShuffleFromZeroables(ResolvedMask, OpUndef, OpZero);
       bool AllowVar = false;
       if (Ops[i].getNode()->hasOneUse() ||
           SDNode::areOnlyUsersOf(CombinedNodes, Ops[i].getNode()))
         AllowVar = AllowVariableMask;
       if (SDValue Res = combineX86ShufflesRecursively(
-              Ops, i, Root, ResolvedMask, CombinedNodes, Depth + 1,
+              Ops, i, Root, ResolvedMask, CombinedNodes, Depth + 1, MaxDepth,
               HasVariableMask, AllowVar, DAG, Subtarget))
         return Res;
     }
   }
 
   // Attempt to constant fold all of the constant source ops.
   if (SDValue Cst = combineX86ShufflesConstants(
           Ops, Mask, Root, HasVariableMask, DAG, Subtarget))

                                            HasVariableMask, AllowVariableMask,
                                            DAG, Subtarget);
 }
 
 /// Helper entry wrapper to combineX86ShufflesRecursively.
 static SDValue combineX86ShufflesRecursively(SDValue Op, SelectionDAG &DAG,
                                              const X86Subtarget &Subtarget) {
   return combineX86ShufflesRecursively({Op}, 0, Op, {0}, {}, /*Depth*/ 0,
+                                       X86::MaxShuffleCombineDepth,
                                        /*HasVarMask*/ false,
                                        /*AllowVarMask*/ true, DAG, Subtarget);
 }
 
 /// Get the PSHUF-style mask from PSHUF node.
 ///
 /// This is a very minor wrapper around getTargetShuffleMask to easy forming v4
 /// PSHUF-style masks that can be reused with such instructions.

         VT.getScalarSizeInBits() % BCVT.getScalarSizeInBits() == 0) {
       unsigned Scale = VT.getScalarSizeInBits() / BCVT.getScalarSizeInBits();
       SmallVector<int, 16> DemandedMask(BCVT.getVectorNumElements(),
                                         SM_SentinelUndef);
       for (unsigned i = 0; i != Scale; ++i)
         DemandedMask[i] = i;
       if (SDValue Res = combineX86ShufflesRecursively(
               {BC}, 0, BC, DemandedMask, {}, /*Depth*/ 0,
+              X86::MaxShuffleCombineDepth,
               /*HasVarMask*/ false, /*AllowVarMask*/ true, DAG, Subtarget))
         return DAG.getNode(X86ISD::VBROADCAST, DL, VT,
                            DAG.getBitcast(SrcVT, Res));
     }
 
     // broadcast(bitcast(src)) -> bitcast(broadcast(src))
     // 32-bit targets have to bitcast i64 to f64, so better to bitcast upward.
     if (Src.getOpcode() == ISD::BITCAST &&

     APInt SrcUndef, SrcZero;
     if (SimplifyDemandedVectorElts(OpInputs[Src], SrcElts, SrcUndef, SrcZero,
                                    TLO, Depth + 1))
       return true;
   }
 
   // If we don't demand all elements, then attempt to combine to a simpler
   // shuffle.
-  // TODO: Handle other depths, but first we need to handle the fact that
-  // it might combine to the same shuffle.
-  if (!DemandedElts.isAllOnesValue() && Depth == 0) {
+  // We need to convert the depth to something combineX86ShufflesRecursively
+  // can handle - so pretend its Depth == 0 again, and reduce the max depth
+  // to match. This prevents combineX86ShuffleChain from returning a
+  // combined shuffle that's the same as the original root, causing an
+  // infinite loop.
+  if (!DemandedElts.isAllOnesValue()) {
+    assert(Depth < X86::MaxShuffleCombineDepth && "Depth out of range");
+
     SmallVector<int, 64> DemandedMask(NumElts, SM_SentinelUndef);
     for (int i = 0; i != NumElts; ++i)
       if (DemandedElts[i])
         DemandedMask[i] = i;
 
     SDValue NewShuffle = combineX86ShufflesRecursively(
-        {Op}, 0, Op, DemandedMask, {}, Depth, /*HasVarMask*/ false,
+        {Op}, 0, Op, DemandedMask, {}, 0, X86::MaxShuffleCombineDepth - Depth,
+        /*HasVarMask*/ false,
         /*AllowVarMask*/ true, TLO.DAG, Subtarget);
     if (NewShuffle)
       return TLO.CombineTo(Op, NewShuffle);
   }
 
   return false;
 }
 

           continue;
         int VecIdx = Scale * Idx + i;
         ShuffleMask[VecIdx] =
             EltBits[i].isNullValue() ? SM_SentinelZero : VecIdx;
       }
 
       if (SDValue Shuffle = combineX86ShufflesRecursively(
               {SrcVec}, 0, SrcVec, ShuffleMask, {}, /*Depth*/ 1,
+              X86::MaxShuffleCombineDepth,
               /*HasVarMask*/ false, /*AllowVarMask*/ true, DAG, Subtarget))
         return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(N), VT, Shuffle,
                            N->getOperand(0).getOperand(1));
     }
   }
 
   return SDValue();
 }

llvm/test/CodeGen/X86/avg.ll

 ; AVX2-NEXT:    vmovq %r12, %xmm4
 ; AVX2-NEXT:    vmovq %rbx, %xmm5
 ; AVX2-NEXT:    vmovq %rax, %xmm6
 ; AVX2-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
 ; AVX2-NEXT:    shrq %rax
 ; AVX2-NEXT:    vmovq %rax, %xmm7
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; AVX2-NEXT:    vpslldq {{.*#+}} xmm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm8[0,1]
+; AVX2-NEXT:    vpbroadcastw %xmm8, %xmm8
 ; AVX2-NEXT:    vpbroadcastw %xmm9, %xmm0
 ; AVX2-NEXT:    vpblendw {{.*#+}} xmm8 = xmm0[0,1,2,3,4,5,6],xmm8[7]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3],xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} xmm9 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3],xmm15[4],xmm14[4],xmm15[5],xmm14[5],xmm15[6],xmm14[6],xmm15[7],xmm14[7]
-; AVX2-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm9, %xmm1
 ; AVX2-NEXT:    vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5],xmm1[6,7]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm8[3]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3],xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
 ; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
 ; AVX2-NEXT:    vpbroadcastw %xmm2, %xmm2
 ; AVX2-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3],xmm2[4,5,6,7]

llvm/test/CodeGen/X86/avx-trunc.ll

; CHECK-NEXT:    retq
   %B = trunc <4 x i64> %A to <4 x i32>
   ret <4 x i32>%B
 }
 
 define <8 x i16> @trunc_32_16(<8 x i32> %A) nounwind uwtable readnone ssp{
 ; CHECK-LABEL: trunc_32_16:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; CHECK-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; CHECK-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; CHECK-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %B = trunc <8 x i32> %A to <8 x i16>
   ret <8 x i16>%B
 }

llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll

 ; X64-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
 ; X64-NEXT:    vpsrlq $32, %xmm1, %xmm3
 ; X64-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; X64-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; X64-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; X64-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; X64-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; X64-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; X64-NEXT:    vpmuludq %xmm0, %xmm2, %xmm2
 ; X64-NEXT:    vpsrlq $32, %xmm0, %xmm3
 ; X64-NEXT:    vpmuludq %xmm3, %xmm1, %xmm3
 ; X64-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; X64-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; X64-NEXT:    vpmuludq %xmm0, %xmm1, %xmm0
 ; X64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; X64-NEXT:    vmovq %xmm0, %rax

 ; X64-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
 ; X64-NEXT:    vpsrlq $32, %xmm1, %xmm3
 ; X64-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; X64-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; X64-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; X64-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; X64-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; X64-NEXT:    vpsrldq {{.*#+}} xmm2 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
 ; X64-NEXT:    vpmuludq %xmm0, %xmm2, %xmm2
 ; X64-NEXT:    vpsrlq $32, %xmm0, %xmm3
 ; X64-NEXT:    vpmuludq %xmm3, %xmm1, %xmm3
 ; X64-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; X64-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; X64-NEXT:    vpmuludq %xmm0, %xmm1, %xmm0
 ; X64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; X64-NEXT:    vmovq %xmm0, %rax

llvm/test/CodeGen/X86/bitcast-and-setcc-128.ll

   ret i16 %res
 }
 
 define i2 @v2i8(<2 x i8> %a, <2 x i8> %b, <2 x i8> %c, <2 x i8> %d) {
 ; SSE2-LABEL: v2i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    pcmpgtb %xmm3, %xmm2
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    movmskpd %xmm1, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: v2i8:
 ; SSSE3:       # %bb.0:

   %res = bitcast <2 x i1> %y to i2
   ret i2 %res
 }
 
 define i2 @v2i16(<2 x i16> %a, <2 x i16> %b, <2 x i16> %c, <2 x i16> %d) {
 ; SSE2-SSSE3-LABEL: v2i16:
 ; SSE2-SSSE3:       # %bb.0:
 ; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-SSSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSE2-SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE2-SSSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,0,2,1,4,5,6,7]
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm1
 ; SSE2-SSSE3-NEXT:    movmskpd %xmm1, %eax
 ; SSE2-SSSE3-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v2i16:
 ; AVX12:       # %bb.0:

llvm/test/CodeGen/X86/bitcast-setcc-128.ll

   ret i16 %res
 }
 
 define i2 @v2i8(<2 x i8> %a, <2 x i8> %b) {
 ; SSE2-LABEL: v2i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-NEXT:    movmskpd %xmm0, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: v2i8:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    pcmpgtb %xmm1, %xmm0

   %res = bitcast <2 x i1> %x to i2
   ret i2 %res
 }
 
 define i2 @v2i16(<2 x i16> %a, <2 x i16> %b) {
 ; SSE2-SSSE3-LABEL: v2i16:
 ; SSE2-SSSE3:       # %bb.0:
 ; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-SSSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE2-SSSE3-NEXT:    movmskpd %xmm0, %eax
 ; SSE2-SSSE3-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v2i16:
 ; AVX12:       # %bb.0:
 ; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0

llvm/test/CodeGen/X86/buildvec-extract.ll

   %z = zext i16 %e to i64
   %r = insertelement <2 x i64> zeroinitializer, i64 %z, i32 0
   ret <2 x i64> %r
 }
 
 define <2 x i64> @extract1_i16_zext_insert0_i64_undef(<8 x i16> %x) {
 ; SSE-LABEL: extract1_i16_zext_insert0_i64_undef:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: extract1_i16_zext_insert0_i64_undef:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3]
+; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX-NEXT:    retq
   %e = extractelement <8 x i16> %x, i32 1
   %z = zext i16 %e to i64
   %r = insertelement <2 x i64> undef, i64 %z, i32 0
   ret <2 x i64> %r
 }
 

   %z = zext i16 %e to i64
   %r = insertelement <2 x i64> zeroinitializer, i64 %z, i32 1
   ret <2 x i64> %r
 }
 
 define <2 x i64> @extract3_i16_zext_insert1_i64_undef(<8 x i16> %x) {
 ; SSE2-LABEL: extract3_i16_zext_insert1_i64_undef:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7]
-; SSE2-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE2-NEXT:    psrlq $48, %xmm0
 ; SSE2-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7]
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: extract3_i16_zext_insert1_i64_undef:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
 ; SSE41-NEXT:    pxor %xmm1, %xmm1
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4],xmm1[5,6,7]

llvm/test/CodeGen/X86/buildvec-insertvec.ll

 }
 
 ; Verify that the DAGCombiner doesn't wrongly fold a build_vector into a
 ; blend with a zero vector if the build_vector contains negative zero.
 
 define <4 x float> @test_negative_zero_1(<4 x float> %A) {
 ; SSE2-LABEL: test_negative_zero_1:
 ; SSE2:       # %bb.0: # %entry
-; SSE2-NEXT:    movaps %xmm0, %xmm1
-; SSE2-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT:    xorps %xmm2, %xmm2
-; SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; SSE2-NEXT:    xorps %xmm1, %xmm1
+; SSE2-NEXT:    movaps %xmm0, %xmm2
+; SSE2-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSE2-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_negative_zero_1:
 ; SSE41:       # %bb.0: # %entry
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2],zero
 ; SSE41-NEXT:    retq
 ;

 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: test_buildvector_v4f32_load:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]

llvm/test/CodeGen/X86/combine-fcopysign.ll

 ; SSE-NEXT:    orps %xmm7, %xmm3
 ; SSE-NEXT:    movshdup {{.*#+}} xmm5 = xmm0[1,1,3,3]
 ; SSE-NEXT:    andps %xmm4, %xmm5
 ; SSE-NEXT:    movhlps {{.*#+}} xmm1 = xmm1[1,1]
 ; SSE-NEXT:    cvtsd2ss %xmm1, %xmm1
 ; SSE-NEXT:    andps {{.*}}(%rip), %xmm1
 ; SSE-NEXT:    orps %xmm5, %xmm1
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    insertps {{.*#+}} xmm3 = xmm3[0,1],xmm6[0],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm6[0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; SSE-NEXT:    andps %xmm4, %xmm0
 ; SSE-NEXT:    movhlps {{.*#+}} xmm2 = xmm2[1,1]
 ; SSE-NEXT:    xorps %xmm1, %xmm1
 ; SSE-NEXT:    cvtsd2ss %xmm2, %xmm1
 ; SSE-NEXT:    andnps %xmm1, %xmm4
 ; SSE-NEXT:    orps %xmm0, %xmm4
 ; SSE-NEXT:    insertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]

llvm/test/CodeGen/X86/combine-shl.ll

   %2 = shl <4 x i32> %1, <i32 5, i32 5, i32 5, i32 5>
   ret <4 x i32> %2
 }
 
 define <4 x i32> @combine_vec_shl_ge_ashr_extact1(<4 x i32> %x) {
 ; SSE2-LABEL: combine_vec_shl_ge_ashr_extact1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrad $5, %xmm1
+; SSE2-NEXT:    psrad $3, %xmm1
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    psrad $3, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,3]
+; SSE2-NEXT:    psrad $5, %xmm2
+; SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrad $8, %xmm1
 ; SSE2-NEXT:    psrad $4, %xmm0
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[3,3]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [32,64,128,256]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]

   %2 = shl <4 x i32> %1, <i32 3, i32 3, i32 3, i32 3>
   ret <4 x i32> %2
 }
 
 define <4 x i32> @combine_vec_shl_lt_ashr_extact1(<4 x i32> %x) {
 ; SSE2-LABEL: combine_vec_shl_lt_ashr_extact1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrad $7, %xmm1
+; SSE2-NEXT:    psrad $5, %xmm1
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    psrad $5, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,3]
+; SSE2-NEXT:    psrad $7, %xmm2
+; SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrad $8, %xmm1
 ; SSE2-NEXT:    psrad $6, %xmm0
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[3,3]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [8,16,32,256]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]

   %2 = shl <4 x i32> %1, <i32 5, i32 5, i32 5, i32 5>
   ret <4 x i32> %2
 }
 
 define <4 x i32> @combine_vec_shl_gt_lshr1(<4 x i32> %x) {
 ; SSE2-LABEL: combine_vec_shl_gt_lshr1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrld $5, %xmm1
+; SSE2-NEXT:    psrld $3, %xmm1
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    psrld $3, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,3]
+; SSE2-NEXT:    psrld $5, %xmm2
+; SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrld $8, %xmm1
 ; SSE2-NEXT:    psrld $4, %xmm0
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[3,3]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [32,64,128,256]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]

   %2 = shl <4 x i32> %1, <i32 3, i32 3, i32 3, i32 3>
   ret <4 x i32> %2
 }
 
 define <4 x i32> @combine_vec_shl_le_lshr1(<4 x i32> %x) {
 ; SSE2-LABEL: combine_vec_shl_le_lshr1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrld $7, %xmm1
+; SSE2-NEXT:    psrld $5, %xmm1
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    psrld $5, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,3]
+; SSE2-NEXT:    psrld $7, %xmm2
+; SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    psrld $8, %xmm1
 ; SSE2-NEXT:    psrld $6, %xmm0
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[3,3]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [8,16,32,256]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]

llvm/test/CodeGen/X86/insert-into-constant-vector.ll

 ; X64AVX-NEXT:    retq
    %ins = insertelement <8 x i16> <i16 42, i16 1, i16 2, i16 3, i16 4, i16 5, i16 6, i16 7>, i16 %x, i32 5
    ret <8 x i16> %ins
 }
 
 define <4 x i32> @elt3_v4i32(i32 %x) {
 ; X32SSE2-LABEL: elt3_v4i32:
 ; X32SSE2:       # %bb.0:
-; X32SSE2-NEXT:    movaps {{.*#+}} xmm0 = <42,1,2,u>
 ; X32SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X32SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; X32SSE2-NEXT:    movaps {{.*#+}} xmm0 = <42,1,2,u>
+; X32SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; X32SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; X32SSE2-NEXT:    retl
 ;
 ; X64SSE2-LABEL: elt3_v4i32:
 ; X64SSE2:       # %bb.0:
 ; X64SSE2-NEXT:    movd %edi, %xmm1
 ; X64SSE2-NEXT:    movaps {{.*#+}} xmm0 = <42,1,2,u>
-; X64SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; X64SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; X64SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; X64SSE2-NEXT:    retq
 ;
 ; X32SSE4-LABEL: elt3_v4i32:
 ; X32SSE4:       # %bb.0:
 ; X32SSE4-NEXT:    movdqa {{.*#+}} xmm0 = <42,1,2,u>
 ; X32SSE4-NEXT:    pinsrd $3, {{[0-9]+}}(%esp), %xmm0
 ; X32SSE4-NEXT:    retl
; X64AVX-NEXT:    retq
    ret <2 x i64> %ins
 }
 
 define <4 x float> @elt1_v4f32(float %x) {
 ; X32SSE2-LABEL: elt1_v4f32:
 ; X32SSE2:       # %bb.0:
 ; X32SSE2-NEXT:    movaps {{.*#+}} xmm1 = <4.2E+1,u,2.0E+0,3.0E+0>
 ; X32SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X32SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; X32SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; X32SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; X32SSE2-NEXT:    retl
 ;
 ; X64SSE2-LABEL: elt1_v4f32:
 ; X64SSE2:       # %bb.0:
 ; X64SSE2-NEXT:    movaps {{.*#+}} xmm1 = <4.2E+1,u,2.0E+0,3.0E+0>
-; X64SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; X64SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; X64SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; X64SSE2-NEXT:    retq
 ;
 ; X32SSE4-LABEL: elt1_v4f32:
 ; X32SSE4:       # %bb.0:
 ; X32SSE4-NEXT:    movaps {{.*#+}} xmm0 = <4.2E+1,u,2.0E+0,3.0E+0>
 ; X32SSE4-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; X32SSE4-NEXT:    retl

 ; X64AVX-NEXT:    retq
    %ins = insertelement <2 x double> <double 42.0, double 1.0>, double %x, i32 1
    ret <2 x double> %ins
 }
 
 define <8 x i32> @elt7_v8i32(i32 %x) {
 ; X32SSE2-LABEL: elt7_v8i32:
 ; X32SSE2:       # %bb.0:
-; X32SSE2-NEXT:    movaps {{.*#+}} xmm1 = <4,5,6,u>
 ; X32SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X32SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; X32SSE2-NEXT:    movaps {{.*#+}} xmm1 = <4,5,6,u>
+; X32SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; X32SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; X32SSE2-NEXT:    movaps {{.*#+}} xmm0 = [42,1,2,3]
 ; X32SSE2-NEXT:    retl
 ;
 ; X64SSE2-LABEL: elt7_v8i32:
 ; X64SSE2:       # %bb.0:
 ; X64SSE2-NEXT:    movd %edi, %xmm0
 ; X64SSE2-NEXT:    movaps {{.*#+}} xmm1 = <4,5,6,u>
-; X64SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; X64SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; X64SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; X64SSE2-NEXT:    movaps {{.*#+}} xmm0 = [42,1,2,3]
 ; X64SSE2-NEXT:    retq
 ;
 ; X32SSE4-LABEL: elt7_v8i32:
 ; X32SSE4:       # %bb.0:
 ; X32SSE4-NEXT:    movdqa {{.*#+}} xmm1 = <4,5,6,u>
 ; X32SSE4-NEXT:    pinsrd $3, {{[0-9]+}}(%esp), %xmm1

llvm/test/CodeGen/X86/insertelement-shuffle.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown   -mattr=avx2    | FileCheck %s --check-prefixes=X86,X86_AVX256
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2    | FileCheck %s --check-prefixes=X64,X64_AVX256
 ; RUN: llc < %s -mtriple=i686-unknown-unknown   -mattr=avx512f | FileCheck %s --check-prefixes=X86,X86_AVX512
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512f | FileCheck %s --check-prefixes=X64,X64_AVX512
 
 define <8 x float> @insert_subvector_256(i16 %x0, i16 %x1, <8 x float> %v) nounwind {
 ; X86-LABEL: insert_subvector_256:
 ; X86:       # %bb.0:
 ; X86-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; X86-NEXT:    vpinsrw $1, {{[0-9]+}}(%esp), %xmm1, %xmm1
-; X86-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; X86-NEXT:    vpbroadcastd %xmm1, %xmm1
 ; X86-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: insert_subvector_256:
 ; X64:       # %bb.0:
 ; X64-NEXT:    vmovd %edi, %xmm1
 ; X64-NEXT:    vpinsrw $1, %esi, %xmm1, %xmm1
-; X64-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; X64-NEXT:    vpbroadcastd %xmm1, %xmm1
 ; X64-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
 ; X64-NEXT:    retq
   %ins1 = insertelement <2 x i16> undef, i16 %x0, i32 0
   %ins2 = insertelement <2 x i16> %ins1, i16 %x1, i32 1
   %bc = bitcast <2 x i16> %ins2 to float
   %ins3 = insertelement <8 x float> %v, float %bc, i32 1
   ret <8 x float> %ins3
 }

llvm/test/CodeGen/X86/known-signbits-vector.ll

 ; X64-AVX2-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; X64-AVX2-NEXT:    retq
   %1 = sext <2 x i16> %a0 to <2 x i64>
   %2 = shl <2 x i64> %1, <i64 11, i64 5>
   %3 = sitofp <2 x i64> %2 to <2 x double>
   ret <2 x double> %3
 }
 
-; TODO: Fix vpshufd+vpsrlq -> vpshufd/vpermilps
 define <2 x double> @signbits_ashr_concat_ashr_extract_sitofp(<2 x i64> %a0, <4 x i64> %a1) nounwind {
-; X86-LABEL: signbits_ashr_concat_ashr_extract_sitofp:
-; X86:       # %bb.0:
-; X86-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3]
-; X86-NEXT:    vcvtdq2pd %xmm0, %xmm0
-; X86-NEXT:    retl
-;
-; X64-LABEL: signbits_ashr_concat_ashr_extract_sitofp:
-; X64:       # %bb.0:
-; X64-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; X64-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; X64-NEXT:    vcvtdq2pd %xmm0, %xmm0
-; X64-NEXT:    retq
+; CHECK-LABEL: signbits_ashr_concat_ashr_extract_sitofp:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; CHECK-NEXT:    vcvtdq2pd %xmm0, %xmm0
+; CHECK-NEXT:    ret{{[l|q]}}
   %1 = ashr <2 x i64> %a0, <i64 16, i64 16>
   %2 = shufflevector <2 x i64> %1, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
   %3 = shufflevector <4 x i64> %a1, <4 x i64> %2, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   %4 = ashr <4 x i64> %3, <i64 16, i64 16, i64 16, i64 16>
   %5 = shufflevector <4 x i64> %4, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
   %6 = sitofp <2 x i64> %5 to <2 x double>
   ret <2 x double> %6
 }

llvm/test/CodeGen/X86/load-partial.ll

   %1 = load <2 x i8>, <2 x i8>* bitcast (i8* @h to <2 x i8>*), align 1
   %2 = shufflevector <2 x i8> %1, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
   %3 = insertelement <4 x i8> %2, i8 2, i32 2
   %4 = bitcast <4 x i8> %3 to i32
   ret i32 %4
 }
 
 define void @PR43227(i32* %explicit_0, <8 x i32>* %explicit_1) {
-; SSE2-LABEL: PR43227:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    xorps %xmm2, %xmm2
-; SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; SSE2-NEXT:    movaps %xmm1, 672(%rsi)
-; SSE2-NEXT:    movaps %xmm2, 688(%rsi)
-; SSE2-NEXT:    retq
+; SSE-LABEL: PR43227:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
+; SSE-NEXT:    movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSE-NEXT:    psrlq $32, %xmm0
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm1, 672(%rsi)
+; SSE-NEXT:    movdqa %xmm0, 688(%rsi)
+; SSE-NEXT:    retq
 ;
-; SSSE3-LABEL: PR43227:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    xorps %xmm2, %xmm2
-; SSSE3-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSSE3-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSSE3-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; SSSE3-NEXT:    movaps %xmm1, 672(%rsi)
-; SSSE3-NEXT:    movaps %xmm2, 688(%rsi)
-; SSSE3-NEXT:    retq
+; AVX1-LABEL: PR43227:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; AVX1-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    vmovaps %ymm0, 672(%rsi)
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
 ;
-; SSE41-LABEL: PR43227:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE41-NEXT:    pxor %xmm1, %xmm1
-; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
-; SSE41-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE41-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE41-NEXT:    movdqa %xmm1, 672(%rsi)
-; SSE41-NEXT:    movdqa %xmm0, 688(%rsi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: PR43227:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
-; AVX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; AVX-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX-NEXT:    vmovaps %ymm0, 672(%rsi)
-; AVX-NEXT:    vzeroupper
-; AVX-NEXT:    retq
+; AVX2-LABEL: PR43227:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; AVX2-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
+; AVX2-NEXT:    vmovdqa %ymm0, 672(%rsi)
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
   %1 = getelementptr i32, i32* %explicit_0, i64 63
   %2 = bitcast i32* %1 to <3 x i32>*
   %3 = load <3 x i32>, <3 x i32>* %2, align 1
   %4 = shufflevector <3 x i32> %3, <3 x i32> undef, <2 x i32> <i32 1, i32 2>
   %5 = shufflevector <2 x i32> %4, <2 x i32> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %6 = shufflevector <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 undef, i32 0>, <8 x i32> %5, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 5, i32 9, i32 7>
   %7 = getelementptr inbounds <8 x i32>, <8 x i32>* %explicit_1, i64 21
   store <8 x i32> %6, <8 x i32>* %7, align 32
   ret void
 }

llvm/test/CodeGen/X86/load-slice.ll

 ;
 ; The layout is:
 ; LSB 0 1 2 3 | 4 5 6 7 MSB
 ;       Low      High
 ; The base address points to 0 and is 8-bytes aligned.
 ; Low slice starts at 0 (base) and is 8-bytes aligned.
 ; High slice starts at 4 (base + 4-bytes) and is 4-bytes aligned.
 ;
-; STRESS-LABEL: t1:
+; STRESS-LABEL: _t1:
 ; Load out[out_start + 8].real, this is base + 8 * 8 + 0.
 ; STRESS: vmovss 64([[BASE:[^(]+]]), [[OUT_Real:%xmm[0-9]+]]
 ; Load out[out_start + 8].imm, this is base + 8 * 8 + 4.
 ; STRESS-NEXT: vmovss 68([[BASE]]), [[OUT_Imm:%xmm[0-9]+]]
 ; Add low slice: out[out_start].real, this is base + 0.
 ; STRESS-NEXT: vaddss ([[BASE]]), [[OUT_Real]], [[RES_Real:%xmm[0-9]+]]
 ; Add high slice: out[out_start].imm, this is base + 4.
 ; STRESS-NEXT: vaddss 4([[BASE]]), [[OUT_Imm]], [[RES_Imm:%xmm[0-9]+]]
 ; Swap Imm and Real.
 ; STRESS-NEXT: vinsertps $16, [[RES_Imm]], [[RES_Real]], [[RES_Vec:%xmm[0-9]+]]
 ; Put the results back into out[out_start].
 ; STRESS-NEXT: vmovlps [[RES_Vec]], ([[BASE]])
 ;
 ; Same for REGULAR, we eliminate register bank copy with each slices.
-; REGULAR-LABEL: t1:
+; REGULAR-LABEL: _t1:
 ; Load out[out_start + 8].real, this is base + 8 * 8 + 0.
 ; REGULAR: vmovss 64([[BASE:[^)]+]]), [[OUT_Real:%xmm[0-9]+]]
 ; Load out[out_start + 8].imm, this is base + 8 * 8 + 4.
 ; REGULAR-NEXT: vmovss 68([[BASE]]), [[OUT_Imm:%xmm[0-9]+]]
 ; Add low slice: out[out_start].real, this is base + 0.
 ; REGULAR-NEXT: vaddss ([[BASE]]), [[OUT_Real]], [[RES_Real:%xmm[0-9]+]]
 ; Add high slice: out[out_start].imm, this is base + 4.
 ; REGULAR-NEXT: vaddss 4([[BASE]]), [[OUT_Imm]], [[RES_Imm:%xmm[0-9]+]]

 ;
 ; The memory layout is:
 ; LSB 0 1 2 3 | 4 5 | 6 7 MSB
 ;      Low            High
 ; The base address points to 0 and is 8-bytes aligned.
 ; Low slice starts at 0 (base) and is 8-bytes aligned.
 ; High slice starts at 6 (base + 6-bytes) and is 2-bytes aligned.
 ;
-; STRESS-LABEL: t2:
+; STRESS-LABEL: _t2:
 ; STRESS: movzwl 6([[BASE:[^)]+]]), %eax
 ; STRESS-NEXT: addl ([[BASE]]), %eax
 ; STRESS-NEXT: ret
 ;
 ; For the REGULAR heuristic, this is not profitable to slice things that are not
 ; next to each other in memory. Here we have a hole with bytes #4-5.
-; REGULAR-LABEL: t2:
+; REGULAR-LABEL: _t2:
 ; REGULAR: shrq $48
 define i32 @t2(%class.Complex* nocapture %out, i64 %out_start) {
   %arrayidx = getelementptr inbounds %class.Complex, %class.Complex* %out, i64 %out_start
   %bitcast = bitcast %class.Complex* %arrayidx to i64*
   %chunk64 = load i64, i64* %bitcast, align 8
   %slice32_low = trunc i64 %chunk64 to i32
   %shift48 = lshr i64 %chunk64, 48
   %slice32_high = trunc i64 %shift48 to i32
   %res = add i32 %slice32_high, %slice32_low
   ret i32 %res
 }
 
 ; Check that we do not optimize overlapping slices.
 ;
 ; The 64-bits should NOT have been split in as slices are overlapping.
 ; First slice uses bytes numbered 0 to 3.
 ; Second slice uses bytes numbered 6 and 7.
 ; Third slice uses bytes numbered 4 to 7.
 ;
-; STRESS-LABEL: t3:
+; STRESS-LABEL: _t3:
 ; STRESS: shrq $48
 ; STRESS: shrq $32
 ;
-; REGULAR-LABEL: t3:
+; REGULAR-LABEL: _t3:
 ; REGULAR: shrq $48
 ; REGULAR: shrq $32
 define i32 @t3(%class.Complex* nocapture %out, i64 %out_start) {
   %arrayidx = getelementptr inbounds %class.Complex, %class.Complex* %out, i64 %out_start
   %bitcast = bitcast %class.Complex* %arrayidx to i64*
   %chunk64 = load i64, i64* %bitcast, align 8
   %slice32_low = trunc i64 %chunk64 to i32
   %shift48 = lshr i64 %chunk64, 48
   %slice32_high = trunc i64 %shift48 to i32
   %shift32 = lshr i64 %chunk64, 32
   %slice32_lowhigh = trunc i64 %shift32 to i32
   %tmpres = add i32 %slice32_high, %slice32_low
   %res = add i32 %slice32_lowhigh, %tmpres
   ret i32 %res
 }

llvm/test/CodeGen/X86/masked_expandload.ll

 ; SSE2-NEXT:  LBB4_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
 ; SSE2-NEXT:    addq $4, %rdi
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB4_4
 ; SSE2-NEXT:  LBB4_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: expandload_v2f32_v2i1:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm2, %xmm2
 ; SSE42-NEXT:    pcmpeqd %xmm1, %xmm2

   ret <2 x float> %res
 }
 
 define <4 x float> @expandload_v4f32_const(float* %base, <4 x float> %src0) {
 ; SSE2-LABEL: expandload_v4f32_const:
 ; SSE2:       ## %bb.0:
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm2[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[0,2]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: expandload_v4f32_const:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE42-NEXT:    blendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE42-NEXT:    blendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
 ; SSE42-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: expandload_v4f32_const:
 ; AVX1OR2:       ## %bb.0:
 ; AVX1OR2-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512F-LABEL: expandload_v4f32_const:
 ; AVX512F:       ## %bb.0:
 ; AVX512F-NEXT:    ## kill: def $xmm0 killed $xmm0 def $zmm0
 ; AVX512F-NEXT:    movw $7, %ax

 
 define <16 x float> @expandload_v16f32_const(float* %base, <16 x float> %src0) {
 ; SSE2-LABEL: expandload_v16f32_const:
 ; SSE2:       ## %bb.0:
 ; SSE2-NEXT:    movups (%rdi), %xmm0
 ; SSE2-NEXT:    movups 16(%rdi), %xmm1
 ; SSE2-NEXT:    movss {{.*#+}} xmm5 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm4 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,0],xmm5[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm5[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm5 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,0],xmm2[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm5[0,2]
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm5 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,0],xmm2[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm2[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm3[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm2[0,2]
 ; SSE2-NEXT:    movaps %xmm4, %xmm2
 ; SSE2-NEXT:    movaps %xmm5, %xmm3
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: expandload_v16f32_const:
 ; SSE42:       ## %bb.0:
-; SSE42-NEXT:    movups (%rdi), %xmm0
-; SSE42-NEXT:    movups 16(%rdi), %xmm1
+; SSE42-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; SSE42-NEXT:    movsd {{.*#+}} xmm1 = mem[0],zero
+; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
+; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
+; SSE42-NEXT:    insertps {{.*#+}} xmm1 = xmm1[0,1],mem[0],xmm1[3]
+; SSE42-NEXT:    insertps {{.*#+}} xmm1 = xmm1[0,1,2],mem[0]
 ; SSE42-NEXT:    movss {{.*#+}} xmm4 = mem[0],zero,zero,zero
-; SSE42-NEXT:    blendps {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
+; SSE42-NEXT:    blendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
 ; SSE42-NEXT:    movss {{.*#+}} xmm4 = mem[0],zero,zero,zero
-; SSE42-NEXT:    blendps {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
+; SSE42-NEXT:    blendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm3 = xmm3[0],mem[0],xmm3[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm3 = xmm3[0,1],mem[0],xmm3[3]
 ; SSE42-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: expandload_v16f32_const:
 ; AVX1OR2:       ## %bb.0:
-; AVX1OR2-NEXT:    vmovups (%rdi), %xmm0
+; AVX1OR2-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; AVX1OR2-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
+; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
+; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
 ; AVX1OR2-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1OR2-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0],xmm1[1,2,3]
+; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
 ; AVX1OR2-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; AVX1OR2-NEXT:    vextractf128 $1, %ymm1, %xmm1
-; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3]
+; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[2,3]
 ; AVX1OR2-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1],mem[0],xmm1[3]
 ; AVX1OR2-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512F-LABEL: expandload_v16f32_const:
 ; AVX512F:       ## %bb.0:
 ; AVX512F-NEXT:    movw $30719, %ax ## imm = 0x77FF

 ; SSE2-NEXT:  LBB8_60: ## %else114
 ; SSE2-NEXT:    testl $1073741824, %ecx ## imm = 0x40000000
 ; SSE2-NEXT:    jne LBB8_61
 ; SSE2-NEXT:  LBB8_62: ## %else118
 ; SSE2-NEXT:    testl $-2147483648, %ecx ## imm = 0x80000000
 ; SSE2-NEXT:    je LBB8_64
 ; SSE2-NEXT:  LBB8_63: ## %cond.load121
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm7[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm7[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,0]
 ; SSE2-NEXT:  LBB8_64: ## %else122
 ; SSE2-NEXT:    movaps %xmm0, (%rax)
 ; SSE2-NEXT:    movaps %xmm1, 16(%rax)
 ; SSE2-NEXT:    movaps %xmm2, 32(%rax)
 ; SSE2-NEXT:    movaps %xmm3, 48(%rax)
 ; SSE2-NEXT:    movaps %xmm4, 64(%rax)
 ; SSE2-NEXT:    movaps %xmm5, 80(%rax)
 ; SSE2-NEXT:    movaps %xmm6, 96(%rax)
 ; SSE2-NEXT:    movaps %xmm7, 112(%rax)
 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB8_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testb $2, %cl
 ; SSE2-NEXT:    je LBB8_4
 ; SSE2-NEXT:  LBB8_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm0[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm0
 ; SSE2-NEXT:    testb $4, %cl
 ; SSE2-NEXT:    je LBB8_6
 ; SSE2-NEXT:  LBB8_5: ## %cond.load5
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testb $8, %cl
 ; SSE2-NEXT:    je LBB8_8
 ; SSE2-NEXT:  LBB8_7: ## %cond.load9
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testb $16, %cl
 ; SSE2-NEXT:    je LBB8_10
 ; SSE2-NEXT:  LBB8_9: ## %cond.load13
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm8[0],xmm1[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testb $32, %cl
 ; SSE2-NEXT:    je LBB8_12
 ; SSE2-NEXT:  LBB8_11: ## %cond.load17
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm1[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm1
 ; SSE2-NEXT:    testb $64, %cl
 ; SSE2-NEXT:    je LBB8_14
 ; SSE2-NEXT:  LBB8_13: ## %cond.load21
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm1[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testb $-128, %cl
 ; SSE2-NEXT:    je LBB8_16
 ; SSE2-NEXT:  LBB8_15: ## %cond.load25
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $256, %ecx ## imm = 0x100
 ; SSE2-NEXT:    je LBB8_18
 ; SSE2-NEXT:  LBB8_17: ## %cond.load29
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm8[0],xmm2[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $512, %ecx ## imm = 0x200
 ; SSE2-NEXT:    je LBB8_20
 ; SSE2-NEXT:  LBB8_19: ## %cond.load33
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm2[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm2[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm2[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm2
 ; SSE2-NEXT:    testl $1024, %ecx ## imm = 0x400
 ; SSE2-NEXT:    je LBB8_22
 ; SSE2-NEXT:  LBB8_21: ## %cond.load37
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm2[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $2048, %ecx ## imm = 0x800
 ; SSE2-NEXT:    je LBB8_24
 ; SSE2-NEXT:  LBB8_23: ## %cond.load41
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm2[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm2[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $4096, %ecx ## imm = 0x1000
 ; SSE2-NEXT:    je LBB8_26
 ; SSE2-NEXT:  LBB8_25: ## %cond.load45
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = xmm8[0],xmm3[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $8192, %ecx ## imm = 0x2000
 ; SSE2-NEXT:    je LBB8_28
 ; SSE2-NEXT:  LBB8_27: ## %cond.load49
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm3[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm3[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm3[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm3
 ; SSE2-NEXT:    testl $16384, %ecx ## imm = 0x4000
 ; SSE2-NEXT:    je LBB8_30
 ; SSE2-NEXT:  LBB8_29: ## %cond.load53
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm3[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $32768, %ecx ## imm = 0x8000
 ; SSE2-NEXT:    je LBB8_32
 ; SSE2-NEXT:  LBB8_31: ## %cond.load57
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm3[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm3[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $65536, %ecx ## imm = 0x10000
 ; SSE2-NEXT:    je LBB8_34
 ; SSE2-NEXT:  LBB8_33: ## %cond.load61
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm4 = xmm8[0],xmm4[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $131072, %ecx ## imm = 0x20000
 ; SSE2-NEXT:    je LBB8_36
 ; SSE2-NEXT:  LBB8_35: ## %cond.load65
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm4[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm4[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm4[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm4
 ; SSE2-NEXT:    testl $262144, %ecx ## imm = 0x40000
 ; SSE2-NEXT:    je LBB8_38
 ; SSE2-NEXT:  LBB8_37: ## %cond.load69
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm4[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $524288, %ecx ## imm = 0x80000
 ; SSE2-NEXT:    je LBB8_40
 ; SSE2-NEXT:  LBB8_39: ## %cond.load73
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm4[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm4[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $1048576, %ecx ## imm = 0x100000
 ; SSE2-NEXT:    je LBB8_42
 ; SSE2-NEXT:  LBB8_41: ## %cond.load77
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm5 = xmm8[0],xmm5[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $2097152, %ecx ## imm = 0x200000
 ; SSE2-NEXT:    je LBB8_44
 ; SSE2-NEXT:  LBB8_43: ## %cond.load81
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm5[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm5[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm5[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm5
 ; SSE2-NEXT:    testl $4194304, %ecx ## imm = 0x400000
 ; SSE2-NEXT:    je LBB8_46
 ; SSE2-NEXT:  LBB8_45: ## %cond.load85
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm5[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $8388608, %ecx ## imm = 0x800000
 ; SSE2-NEXT:    je LBB8_48
 ; SSE2-NEXT:  LBB8_47: ## %cond.load89
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm5[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $16777216, %ecx ## imm = 0x1000000
 ; SSE2-NEXT:    je LBB8_50
 ; SSE2-NEXT:  LBB8_49: ## %cond.load93
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm6 = xmm8[0],xmm6[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $33554432, %ecx ## imm = 0x2000000
 ; SSE2-NEXT:    je LBB8_52
 ; SSE2-NEXT:  LBB8_51: ## %cond.load97
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm6[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm6[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm6[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm6
 ; SSE2-NEXT:    testl $67108864, %ecx ## imm = 0x4000000
 ; SSE2-NEXT:    je LBB8_54
 ; SSE2-NEXT:  LBB8_53: ## %cond.load101
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm6[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[0,2]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $134217728, %ecx ## imm = 0x8000000
 ; SSE2-NEXT:    je LBB8_56
 ; SSE2-NEXT:  LBB8_55: ## %cond.load105
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm6[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm6[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,0]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $268435456, %ecx ## imm = 0x10000000
 ; SSE2-NEXT:    je LBB8_58
 ; SSE2-NEXT:  LBB8_57: ## %cond.load109
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm7 = xmm8[0],xmm7[1,2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    testl $536870912, %ecx ## imm = 0x20000000
 ; SSE2-NEXT:    je LBB8_60
 ; SSE2-NEXT:  LBB8_59: ## %cond.load113
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm7[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm7[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm7[2,3]
 ; SSE2-NEXT:    addq $4, %rsi
 ; SSE2-NEXT:    movaps %xmm8, %xmm7
 ; SSE2-NEXT:    testl $1073741824, %ecx ## imm = 0x40000000
 ; SSE2-NEXT:    je LBB8_62
 ; SSE2-NEXT:  LBB8_61: ## %cond.load117
 ; SSE2-NEXT:    movss {{.*#+}} xmm8 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,0],xmm7[3,0]

 ; SSE2-NEXT:  LBB10_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
 ; SSE2-NEXT:    addq $4, %rdi
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB10_4
 ; SSE2-NEXT:  LBB10_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; SSE2-NEXT:    addq $4, %rdi
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB10_6
 ; SSE2-NEXT:  LBB10_5: ## %cond.load5
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE2-NEXT:    addq $4, %rdi
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB10_8
 ; SSE2-NEXT:  LBB10_7: ## %cond.load9
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: expandload_v4i32_v4i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm2, %xmm2
 ; SSE42-NEXT:    pcmpeqd %xmm1, %xmm2
 ; SSE42-NEXT:    movmskps %xmm2, %eax

llvm/test/CodeGen/X86/masked_load.ll

 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB7_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB7_4
 ; SSE2-NEXT:  LBB7_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm1
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v2f32_v2i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm2, %xmm2

 ; SSE2-NEXT:  LBB8_4: ## %else2
 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB8_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB8_4
 ; SSE2-NEXT:  LBB8_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v2f32_v2i32_undef:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm1, %xmm1
 ; SSE42-NEXT:    pcmpeqd %xmm0, %xmm1

 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB9_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB9_4
 ; SSE2-NEXT:  LBB9_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm1
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB9_6
 ; SSE2-NEXT:  LBB9_5: ## %cond.load4
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB9_8
 ; SSE2-NEXT:  LBB9_7: ## %cond.load7
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v4f32_v4i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm2, %xmm2
 ; SSE42-NEXT:    pcmpeqd %xmm0, %xmm2

 ; SSE2-NEXT:  LBB10_16: ## %else20
 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB10_1: ## %cond.load
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB10_4
 ; SSE2-NEXT:  LBB10_3: ## %cond.load1
-; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[0,0]
+; SSE2-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[2,3]
 ; SSE2-NEXT:    movaps %xmm2, %xmm0
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB10_6
 ; SSE2-NEXT:  LBB10_5: ## %cond.load4
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0,2]
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB10_8
 ; SSE2-NEXT:  LBB10_7: ## %cond.load7
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
 ; SSE2-NEXT:    testb $16, %al
 ; SSE2-NEXT:    je LBB10_10
 ; SSE2-NEXT:  LBB10_9: ## %cond.load10
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $32, %al
 ; SSE2-NEXT:    je LBB10_12
 ; SSE2-NEXT:  LBB10_11: ## %cond.load13
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm2, %xmm1
 ; SSE2-NEXT:    testb $64, %al
 ; SSE2-NEXT:    je LBB10_14
 ; SSE2-NEXT:  LBB10_13: ## %cond.load16
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm1[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
 ; SSE2-NEXT:    testb $-128, %al
 ; SSE2-NEXT:    je LBB10_16
 ; SSE2-NEXT:  LBB10_15: ## %cond.load19
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v8f32_v8i1_zero:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    psllw $15, %xmm0
 ; SSE42-NEXT:    packsswb %xmm0, %xmm0
 ; SSE42-NEXT:    pmovmskb %xmm0, %eax

 ; SSE2-NEXT:  LBB11_12: ## %else14
 ; SSE2-NEXT:    testb $64, %al
 ; SSE2-NEXT:    jne LBB11_13
 ; SSE2-NEXT:  LBB11_14: ## %else17
 ; SSE2-NEXT:    testb $-128, %al
 ; SSE2-NEXT:    je LBB11_16
 ; SSE2-NEXT:  LBB11_15: ## %cond.load19
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm3[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
 ; SSE2-NEXT:  LBB11_16: ## %else20
 ; SSE2-NEXT:    movaps %xmm2, %xmm0
 ; SSE2-NEXT:    movaps %xmm3, %xmm1
 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB11_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB11_4
 ; SSE2-NEXT:  LBB11_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm2
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB11_6
 ; SSE2-NEXT:  LBB11_5: ## %cond.load4
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB11_8
 ; SSE2-NEXT:  LBB11_7: ## %cond.load7
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
 ; SSE2-NEXT:    testb $16, %al
 ; SSE2-NEXT:    je LBB11_10
 ; SSE2-NEXT:  LBB11_9: ## %cond.load10
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = xmm0[0],xmm3[1,2,3]
 ; SSE2-NEXT:    testb $32, %al
 ; SSE2-NEXT:    je LBB11_12
 ; SSE2-NEXT:  LBB11_11: ## %cond.load13
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm3[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm3
 ; SSE2-NEXT:    testb $64, %al
 ; SSE2-NEXT:    je LBB11_14
 ; SSE2-NEXT:  LBB11_13: ## %cond.load16
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm3[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[0,2]

 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB17_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB17_4
 ; SSE2-NEXT:  LBB17_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm1
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v2i32_v2i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm2, %xmm2

 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB18_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB18_4
 ; SSE2-NEXT:  LBB18_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm1
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB18_6
 ; SSE2-NEXT:  LBB18_5: ## %cond.load4
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB18_8
 ; SSE2-NEXT:  LBB18_7: ## %cond.load7
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v4i32_v4i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pxor %xmm2, %xmm2
 ; SSE42-NEXT:    pcmpeqd %xmm0, %xmm2

 ; SSE2-NEXT:  LBB19_12: ## %else14
 ; SSE2-NEXT:    testb $64, %al
 ; SSE2-NEXT:    jne LBB19_13
 ; SSE2-NEXT:  LBB19_14: ## %else17
 ; SSE2-NEXT:    testb $-128, %al
 ; SSE2-NEXT:    je LBB19_16
 ; SSE2-NEXT:  LBB19_15: ## %cond.load19
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
 ; SSE2-NEXT:  LBB19_16: ## %else20
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    movaps %xmm2, %xmm1
 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB19_1: ## %cond.load
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB19_4
 ; SSE2-NEXT:  LBB19_3: ## %cond.load1
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm1
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB19_6
 ; SSE2-NEXT:  LBB19_5: ## %cond.load4
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB19_8
 ; SSE2-NEXT:  LBB19_7: ## %cond.load7
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE2-NEXT:    testb $16, %al
 ; SSE2-NEXT:    je LBB19_10
 ; SSE2-NEXT:  LBB19_9: ## %cond.load10
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
 ; SSE2-NEXT:    testb $32, %al
 ; SSE2-NEXT:    je LBB19_12
 ; SSE2-NEXT:  LBB19_11: ## %cond.load13
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[2,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm2
 ; SSE2-NEXT:    testb $64, %al
 ; SSE2-NEXT:    je LBB19_14
 ; SSE2-NEXT:  LBB19_13: ## %cond.load16
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]

 ; SSE2-NEXT:  LBB20_16: ## %else20
 ; SSE2-NEXT:    retq
 ; SSE2-NEXT:  LBB20_1: ## %cond.load
 ; SSE2-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
 ; SSE2-NEXT:    testb $2, %al
 ; SSE2-NEXT:    je LBB20_4
 ; SSE2-NEXT:  LBB20_3: ## %cond.load1
-; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[0,0]
+; SSE2-NEXT:    movd {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[2,3]
 ; SSE2-NEXT:    movaps %xmm2, %xmm0
 ; SSE2-NEXT:    testb $4, %al
 ; SSE2-NEXT:    je LBB20_6
 ; SSE2-NEXT:  LBB20_5: ## %cond.load4
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0,2]
 ; SSE2-NEXT:    testb $8, %al
 ; SSE2-NEXT:    je LBB20_8
 ; SSE2-NEXT:  LBB20_7: ## %cond.load7
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
 ; SSE2-NEXT:    testb $16, %al
 ; SSE2-NEXT:    je LBB20_10
 ; SSE2-NEXT:  LBB20_9: ## %cond.load10
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
 ; SSE2-NEXT:    testb $32, %al
 ; SSE2-NEXT:    je LBB20_12
 ; SSE2-NEXT:  LBB20_11: ## %cond.load13
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm1[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,3]
 ; SSE2-NEXT:    movaps %xmm2, %xmm1
 ; SSE2-NEXT:    testb $64, %al
 ; SSE2-NEXT:    je LBB20_14
 ; SSE2-NEXT:  LBB20_13: ## %cond.load16
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm1[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
 ; SSE2-NEXT:    testb $-128, %al
 ; SSE2-NEXT:    je LBB20_16
 ; SSE2-NEXT:  LBB20_15: ## %cond.load19
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: load_v8i32_v8i1_zero:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    psllw $15, %xmm0
 ; SSE42-NEXT:    packsswb %xmm0, %xmm0
 ; SSE42-NEXT:    pmovmskb %xmm0, %eax

 ; 128-bit FP vectors are supported with AVX.
 
 define <4 x float> @mload_constmask_v4f32(<4 x float>* %addr, <4 x float> %dst) {
 ; SSE2-LABEL: mload_constmask_v4f32:
 ; SSE2:       ## %bb.0:
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
+; SSE2-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE2-NEXT:    movaps %xmm0, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
-; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,0],xmm1[2,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: mload_constmask_v4f32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE42-NEXT:    blendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE42-NEXT:    blendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
 ; SSE42-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: mload_constmask_v4f32:
 ; AVX1OR2:       ## %bb.0:
 ; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm0 = mem[0],xmm0[1],mem[2,3]
 ; AVX1OR2-NEXT:    retq

 }
 
 ; 128-bit integer vectors are supported with AVX2.
 
 define <4 x i32> @mload_constmask_v4i32(<4 x i32>* %addr, <4 x i32> %dst) {
 ; SSE2-LABEL: mload_constmask_v4i32:
 ; SSE2:       ## %bb.0:
 ; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
-; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    movaps %xmm1, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[0,2]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[2,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,0]
+; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,0]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: mload_constmask_v4i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pinsrd $1, 4(%rdi), %xmm0
 ; SSE42-NEXT:    pinsrd $2, 8(%rdi), %xmm0
 ; SSE42-NEXT:    pinsrd $3, 12(%rdi), %xmm0

 
 ; 256-bit FP vectors are supported with AVX.
 
 define <8 x float> @mload_constmask_v8f32(<8 x float>* %addr, <8 x float> %dst) {
 ; SSE2-LABEL: mload_constmask_v8f32:
 ; SSE2:       ## %bb.0:
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm3[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[0,2]
 ; SSE2-NEXT:    movaps %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: mload_constmask_v8f32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE42-NEXT:    blendps {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
+; SSE42-NEXT:    blendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]
 ; SSE42-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]
 ; SSE42-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: mload_constmask_v8f32:
 ; AVX1OR2:       ## %bb.0:
 ; AVX1OR2-NEXT:    vmovaps {{.*#+}} ymm1 = [4294967295,4294967295,4294967295,0,0,0,0,0]
 ; AVX1OR2-NEXT:    vmaskmovps (%rdi), %ymm1, %ymm1

 
 ; 256-bit integer vectors are supported with AVX2.
 
 define <8 x i32> @mload_constmask_v8i32(<8 x i32>* %addr, <8 x i32> %dst) {
 ; SSE2-LABEL: mload_constmask_v8i32:
 ; SSE2:       ## %bb.0:
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,0],xmm3[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
 ; SSE2-NEXT:    movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,0],xmm0[3,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[0,2]
 ; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE2-NEXT:    movaps %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: mload_constmask_v8i32:
 ; SSE42:       ## %bb.0:
 ; SSE42-NEXT:    pinsrd $0, (%rdi), %xmm0
 ; SSE42-NEXT:    pinsrd $1, 4(%rdi), %xmm0

llvm/test/CodeGen/X86/masked_store_trunc.ll

 ; SSE4-NEXT:    je .LBB11_16
 ; SSE4-NEXT:  .LBB11_15: # %cond.store13
 ; SSE4-NEXT:    pextrw $7, %xmm0, 14(%rdi)
 ; SSE4-NEXT:    retq
 ;
 ; AVX1-LABEL: truncstore_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1

llvm/test/CodeGen/X86/oddshuffles.ll

   store <12 x i16> %r, <12 x i16>* %p
   ret void
 }
 
 define void @v12i32(<8 x i32> %a, <8 x i32> %b, <12 x i32>* %p) nounwind {
 ; SSE2-LABEL: v12i32:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movaps %xmm2, %xmm3
-; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,0],xmm0[1,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[1,3]
 ; SSE2-NEXT:    movaps %xmm0, %xmm4
 ; SSE2-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
 ; SSE2-NEXT:    movaps %xmm0, %xmm3
-; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[2,1]
+; SSE2-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
 ; SSE2-NEXT:    movaps %xmm2, %xmm5
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm1[1,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[0,2]
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,2],xmm2[3,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm2[3,3]
 ; SSE2-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[0,2]
 ; SSE2-NEXT:    movaps %xmm2, 32(%rdi)
 ; SSE2-NEXT:    movaps %xmm5, 16(%rdi)
 ; SSE2-NEXT:    movaps %xmm4, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: v12i32:

 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    pand %xmm1, %xmm4
 ; SSE2-NEXT:    pandn %xmm2, %xmm1
 ; SSE2-NEXT:    por %xmm4, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,7,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,7,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm8[0,1,2,1]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,0],xmm4[2,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,0]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE2-NEXT:    movdqa %xmm4, %xmm5
 ; SSE2-NEXT:    pandn %xmm2, %xmm5
 ; SSE2-NEXT:    movdqa %xmm3, %xmm6

   store <24 x i16> %interleaved, <24 x i16>* %p, align 4
   ret void
 }
 
 define void @interleave_24i32_out(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2, <8 x i32>* %q3) nounwind {
 ; SSE2-LABEL: interleave_24i32_out:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movups 80(%rdi), %xmm8
-; SSE2-NEXT:    movups 64(%rdi), %xmm4
-; SSE2-NEXT:    movdqu (%rdi), %xmm0
-; SSE2-NEXT:    movups 16(%rdi), %xmm6
+; SSE2-NEXT:    movups 64(%rdi), %xmm3
+; SSE2-NEXT:    movdqu (%rdi), %xmm1
+; SSE2-NEXT:    movups 16(%rdi), %xmm5
 ; SSE2-NEXT:    movups 32(%rdi), %xmm10
-; SSE2-NEXT:    movups 48(%rdi), %xmm12
-; SSE2-NEXT:    movdqa %xmm0, %xmm11
-; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[2,3,2,3]
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm6[0,0]
-; SSE2-NEXT:    movaps %xmm0, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[1,1,1,1]
-; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm10[1,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,3],xmm6[0,2]
-; SSE2-NEXT:    movaps %xmm12, %xmm6
-; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,0],xmm4[0,0]
-; SSE2-NEXT:    movaps %xmm6, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[3,3]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,1,1]
-; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm8[1,0]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[2,3,2,3]
-; SSE2-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm4[0,2]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm8[0,3]
-; SSE2-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,2],xmm2[2,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm8[2,0]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
-; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm10[0,3]
-; SSE2-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,2],xmm1[2,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm10[2,0]
-; SSE2-NEXT:    movups %xmm12, 16(%rsi)
+; SSE2-NEXT:    movdqu 48(%rdi), %xmm2
+; SSE2-NEXT:    movdqa %xmm1, %xmm11
+; SSE2-NEXT:    movaps %xmm10, %xmm7
+; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,1],xmm5[3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm5[0,0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[1,1,1,1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,3],xmm10[1,1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,3],xmm5[0,2]
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    movaps %xmm8, %xmm4
+; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,1],xmm3[3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,3,2,3]
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[0,0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm12 = xmm3[1,1,1,1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,3],xmm8[1,1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[2,0]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm10[0,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[0,3]
+; SSE2-NEXT:    movups %xmm5, 16(%rsi)
 ; SSE2-NEXT:    movups %xmm11, (%rsi)
-; SSE2-NEXT:    movups %xmm6, 16(%rdx)
-; SSE2-NEXT:    movups %xmm0, (%rdx)
-; SSE2-NEXT:    movups %xmm5, 16(%rcx)
-; SSE2-NEXT:    movups %xmm7, (%rcx)
+; SSE2-NEXT:    movups %xmm2, 16(%rdx)
+; SSE2-NEXT:    movups %xmm1, (%rdx)
+; SSE2-NEXT:    movups %xmm6, 16(%rcx)
+; SSE2-NEXT:    movups %xmm0, (%rcx)
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: interleave_24i32_out:
 ; SSE42:       # %bb.0:
 ; SSE42-NEXT:    movups 80(%rdi), %xmm8
 ; SSE42-NEXT:    movdqu 64(%rdi), %xmm9
 ; SSE42-NEXT:    movdqu (%rdi), %xmm4
 ; SSE42-NEXT:    movdqu 16(%rdi), %xmm2

 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movups (%rsi), %xmm1
 ; SSE2-NEXT:    movups 16(%rsi), %xmm0
 ; SSE2-NEXT:    movups (%rdx), %xmm8
 ; SSE2-NEXT:    movups 16(%rdx), %xmm5
 ; SSE2-NEXT:    movups (%rcx), %xmm3
 ; SSE2-NEXT:    movups 16(%rcx), %xmm6
 ; SSE2-NEXT:    movaps %xmm3, %xmm7
-; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,0],xmm1[1,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[1,3]
 ; SSE2-NEXT:    movaps %xmm1, %xmm9
 ; SSE2-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm7[0,2]
 ; SSE2-NEXT:    movaps %xmm5, %xmm7
-; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,2],xmm6[3,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
 ; SSE2-NEXT:    movaps %xmm6, %xmm4
 ; SSE2-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm7[0,2]
 ; SSE2-NEXT:    movaps %xmm0, %xmm7
-; SSE2-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[2,1]
+; SSE2-NEXT:    unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm5[1]
 ; SSE2-NEXT:    movaps %xmm6, %xmm2
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm5[1,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm7[0,2]
-; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,0],xmm0[1,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[1,3]
 ; SSE2-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[0,2]
 ; SSE2-NEXT:    movaps %xmm8, %xmm5
-; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,2],xmm3[3,2]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm3[3,3]
 ; SSE2-NEXT:    movaps %xmm3, %xmm6
 ; SSE2-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm5[0,2]
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm8[2,1]
+; SSE2-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm8[1]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm8[1,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[0,2]
 ; SSE2-NEXT:    movups %xmm3, 16(%rdi)
 ; SSE2-NEXT:    movups %xmm6, 32(%rdi)
 ; SSE2-NEXT:    movups %xmm0, 48(%rdi)
 ; SSE2-NEXT:    movups %xmm2, 64(%rdi)
 ; SSE2-NEXT:    movups %xmm4, 80(%rdi)
 ; SSE2-NEXT:    movups %xmm9, (%rdi)

 ; SSE42-NEXT:    movdqu %xmm0, 64(%rdi)
 ; SSE42-NEXT:    movdqu %xmm7, 80(%rdi)
 ; SSE42-NEXT:    movdqu %xmm1, (%rdi)
 ; SSE42-NEXT:    retq
 ;
 ; AVX1-LABEL: interleave_24i32_in:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovups (%rdx), %xmm0
-; AVX1-NEXT:    vmovups 16(%rdx), %xmm1
-; AVX1-NEXT:    vmovups (%rsi), %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[2,0],xmm0[2,0]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm0[1,1],xmm3[0,2]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm2[0,0]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[2,1]
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-NEXT:    vbroadcastsd (%rcx), %ymm2
-; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7]
-; AVX1-NEXT:    vmovups 16(%rcx), %xmm2
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[3,0],xmm2[3,0]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[2,1],xmm3[0,2]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,0],xmm1[1,0]
-; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,2]
+; AVX1-NEXT:    vmovups (%rsi), %xmm1
+; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm1[1],xmm0[1]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm2[0,2]
+; AVX1-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vbroadcastsd (%rcx), %ymm1
+; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX1-NEXT:    vmovups 16(%rcx), %xmm1
+; AVX1-NEXT:    vmovups 16(%rdx), %xmm2
+; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[3,0],xmm1[3,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[2,1],xmm3[0,2]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[1,0]
+; AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,2]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
 ; AVX1-NEXT:    vbroadcastsd 24(%rsi), %ymm2
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm2 = mem[1,0,2,2]
 ; AVX1-NEXT:    vpermilpd {{.*#+}} ymm3 = mem[1,1,2,2]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6],ymm2[7]
 ; AVX1-NEXT:    vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7]
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]

 ; AVX2-FAST-NEXT:    retq
 ;
 ; XOP-LABEL: interleave_24i32_in:
 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vmovups (%rsi), %ymm0
 ; XOP-NEXT:    vmovups (%rcx), %ymm1
 ; XOP-NEXT:    vpermil2ps {{.*#+}} ymm0 = ymm1[2],ymm0[3],ymm1[u,3],ymm0[4],ymm1[u,4],ymm0[5]
 ; XOP-NEXT:    vmovups (%rdx), %xmm1
-; XOP-NEXT:    vmovups 16(%rdx), %xmm2
-; XOP-NEXT:    vmovups (%rsi), %xmm3
-; XOP-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[2,0],xmm1[2,0]
-; XOP-NEXT:    vshufps {{.*#+}} xmm4 = xmm1[1,1],xmm4[0,2]
-; XOP-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[0,0],xmm3[0,0]
-; XOP-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,1]
-; XOP-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; XOP-NEXT:    vbroadcastsd (%rcx), %ymm3
-; XOP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3,4],ymm3[5],ymm1[6,7]
-; XOP-NEXT:    vmovups 16(%rcx), %xmm3
-; XOP-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[3,0],xmm3[3,0]
-; XOP-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[2,1],xmm4[0,2]
-; XOP-NEXT:    vshufps {{.*#+}} xmm3 = xmm3[1,0],xmm2[1,0]
-; XOP-NEXT:    vshufps {{.*#+}} xmm2 = xmm3[2,0],xmm2[2,2]
+; XOP-NEXT:    vmovups (%rsi), %xmm2
+; XOP-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm1[1]
+; XOP-NEXT:    vshufps {{.*#+}} xmm3 = xmm1[1,1],xmm3[0,2]
+; XOP-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; XOP-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,1]
+; XOP-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; XOP-NEXT:    vbroadcastsd (%rcx), %ymm2
+; XOP-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7]
+; XOP-NEXT:    vmovups 16(%rcx), %xmm2
+; XOP-NEXT:    vmovups 16(%rdx), %xmm3
+; XOP-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[3,0],xmm2[3,0]
+; XOP-NEXT:    vshufps {{.*#+}} xmm4 = xmm2[2,1],xmm4[0,2]
+; XOP-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[1,0]
+; XOP-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[2,2]
 ; XOP-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; XOP-NEXT:    vbroadcastsd 24(%rsi), %ymm3
 ; XOP-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7]
 ; XOP-NEXT:    vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7]
 ; XOP-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4],ymm3[5],ymm0[6,7]
 ; XOP-NEXT:    vmovups %ymm0, 32(%rdi)
 ; XOP-NEXT:    vmovups %ymm2, 64(%rdi)
 ; XOP-NEXT:    vmovups %ymm1, (%rdi)

llvm/test/CodeGen/X86/oddsubvector.ll

 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    shrq $48, %rcx
 ; SSE-NEXT:    movb %cl, 6(%rdi)
 ; SSE-NEXT:    shrl $16, %eax
 ; SSE-NEXT:    movw %ax, 4(%rdi)
 ; SSE-NEXT:    movd %xmm1, (%rdi)
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: insert_v7i8_v2i16_2:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    movl (%rsi), %eax
-; AVX1-NEXT:    vmovd %eax, %xmm0
-; AVX1-NEXT:    movq (%rdi), %rcx
-; AVX1-NEXT:    vmovq %rcx, %xmm1
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-NEXT:    shrq $48, %rcx
-; AVX1-NEXT:    movb %cl, 6(%rdi)
-; AVX1-NEXT:    shrl $16, %eax
-; AVX1-NEXT:    movw %ax, 4(%rdi)
-; AVX1-NEXT:    vmovd %xmm0, (%rdi)
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: insert_v7i8_v2i16_2:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    movl (%rsi), %eax
-; AVX2-NEXT:    vmovd %eax, %xmm0
-; AVX2-NEXT:    movq (%rdi), %rcx
-; AVX2-NEXT:    vmovq %rcx, %xmm1
-; AVX2-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX2-NEXT:    shrq $48, %rcx
-; AVX2-NEXT:    movb %cl, 6(%rdi)
-; AVX2-NEXT:    shrl $16, %eax
-; AVX2-NEXT:    movw %ax, 4(%rdi)
-; AVX2-NEXT:    vmovd %xmm0, (%rdi)
-; AVX2-NEXT:    retq
+; AVX-LABEL: insert_v7i8_v2i16_2:
+; AVX:       # %bb.0:
+; AVX-NEXT:    movl (%rsi), %eax
+; AVX-NEXT:    vmovd %eax, %xmm0
+; AVX-NEXT:    movq (%rdi), %rcx
+; AVX-NEXT:    vmovq %rcx, %xmm1
+; AVX-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX-NEXT:    shrq $48, %rcx
+; AVX-NEXT:    movb %cl, 6(%rdi)
+; AVX-NEXT:    shrl $16, %eax
+; AVX-NEXT:    movw %ax, 4(%rdi)
+; AVX-NEXT:    vmovd %xmm0, (%rdi)
+; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: insert_v7i8_v2i16_2:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    movl (%rsi), %eax
 ; AVX512-NEXT:    vmovd %eax, %xmm0
 ; AVX512-NEXT:    movq (%rdi), %rcx
 ; AVX512-NEXT:    vmovq %rcx, %xmm1
 ; AVX512-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; AVX512-NEXT:    shrq $48, %rcx
 ; AVX512-NEXT:    movb %cl, 6(%rdi)
 ; AVX512-NEXT:    shrl $16, %eax
 ; AVX512-NEXT:    movw %ax, 4(%rdi)
 ; AVX512-NEXT:    vmovd %xmm0, (%rdi)
 ; AVX512-NEXT:    retq
-;
-; XOP-LABEL: insert_v7i8_v2i16_2:
-; XOP:       # %bb.0:
-; XOP-NEXT:    movl (%rsi), %eax
-; XOP-NEXT:    vmovd %eax, %xmm0
-; XOP-NEXT:    movq (%rdi), %rcx
-; XOP-NEXT:    vmovq %rcx, %xmm1
-; XOP-NEXT:    insertq {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,1,2,3],xmm1[6,7,u,u,u,u,u,u,u,u]
-; XOP-NEXT:    shrq $48, %rcx
-; XOP-NEXT:    movb %cl, 6(%rdi)
-; XOP-NEXT:    shrl $16, %eax
-; XOP-NEXT:    movw %ax, 4(%rdi)
-; XOP-NEXT:    vmovd %xmm1, (%rdi)
-; XOP-NEXT:    retq
   %1 = load <2 x i16>, <2 x i16> *%a1
   %2 = bitcast <2 x i16> %1 to <4 x i8>
   %3 = shufflevector <4 x i8> %2, <4 x i8> undef, <7 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef>
   %4 = load <7 x i8>, <7 x i8> *%a0
   %5 = shufflevector <7 x i8> %4, <7 x i8> %3, <7 x i32> <i32 0, i32 1, i32 7, i32 8, i32 9, i32 10, i32 6>
   store <7 x i8> %5, <7 x i8>* %a0
   ret void
 }

llvm/test/CodeGen/X86/pmul.ll

   %rescast = bitcast <8 x i64> %res64 to <16 x i32>
   %res = shufflevector <16 x i32> %rescast, <16 x i32> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7,i32 9, i32 11, i32 13, i32 15 >
   ret <8 x i32> %res
 }
 
 define <8 x i64> @mul_v8i64_sext(<8 x i16> %val1, <8 x i32> %val2) {
 ; SSE2-LABEL: mul_v8i64_sext:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm12
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7]
-; SSE2-NEXT:    psrad $16, %xmm10
-; SSE2-NEXT:    pxor %xmm8, %xmm8
+; SSE2-NEXT:    movdqa %xmm1, %xmm15
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm0[4],xmm14[5],xmm0[5],xmm14[6],xmm0[6],xmm14[7],xmm0[7]
+; SSE2-NEXT:    psrad $16, %xmm14
 ; SSE2-NEXT:    pxor %xmm13, %xmm13
-; SSE2-NEXT:    pcmpgtd %xmm10, %xmm13
-; SSE2-NEXT:    movdqa %xmm10, %xmm9
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm13[2],xmm9[3],xmm13[3]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm13[0],xmm10[1],xmm13[1]
+; SSE2-NEXT:    pxor %xmm10, %xmm10
+; SSE2-NEXT:    pcmpgtd %xmm14, %xmm10
+; SSE2-NEXT:    movdqa %xmm14, %xmm8
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm10[2],xmm8[3],xmm10[3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm10[0],xmm14[1],xmm10[1]
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    pxor %xmm15, %xmm15
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm15
+; SSE2-NEXT:    pxor %xmm5, %xmm5
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
 ; SSE2-NEXT:    movdqa %xmm0, %xmm11
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm15[2],xmm11[3],xmm15[3]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm5[2],xmm11[3],xmm5[3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; SSE2-NEXT:    pxor %xmm14, %xmm14
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm14
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1]
-; SSE2-NEXT:    pxor %xmm5, %xmm5
-; SSE2-NEXT:    pcmpgtd %xmm2, %xmm5
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; SSE2-NEXT:    pxor %xmm9, %xmm9
+; SSE2-NEXT:    pcmpgtd %xmm3, %xmm9
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
+; SSE2-NEXT:    pxor %xmm12, %xmm12
+; SSE2-NEXT:    pcmpgtd %xmm2, %xmm12
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE2-NEXT:    pxor %xmm6, %xmm6
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm6
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
 ; SSE2-NEXT:    pxor %xmm7, %xmm7
-; SSE2-NEXT:    pcmpgtd %xmm12, %xmm7
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm7[0],xmm12[1],xmm7[1]
-; SSE2-NEXT:    movdqa %xmm15, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm12, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm0, %xmm7
-; SSE2-NEXT:    paddq %xmm4, %xmm7
-; SSE2-NEXT:    psllq $32, %xmm7
-; SSE2-NEXT:    pmuludq %xmm12, %xmm0
-; SSE2-NEXT:    paddq %xmm7, %xmm0
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm8[2],xmm15[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm1, %xmm15
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm11, %xmm6
-; SSE2-NEXT:    paddq %xmm15, %xmm6
-; SSE2-NEXT:    psllq $32, %xmm6
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm7
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
+; SSE2-NEXT:    pcmpgtd %xmm15, %xmm13
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm13[0],xmm15[1],xmm13[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm15, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm4
+; SSE2-NEXT:    paddq %xmm6, %xmm4
+; SSE2-NEXT:    psllq $32, %xmm4
+; SSE2-NEXT:    pmuludq %xmm15, %xmm0
+; SSE2-NEXT:    paddq %xmm4, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm1, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm11, %xmm5
+; SSE2-NEXT:    paddq %xmm4, %xmm5
+; SSE2-NEXT:    psllq $32, %xmm5
 ; SSE2-NEXT:    pmuludq %xmm11, %xmm1
-; SSE2-NEXT:    paddq %xmm6, %xmm1
-; SSE2-NEXT:    movdqa %xmm13, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
+; SSE2-NEXT:    paddq %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[0,1,1,3]
 ; SSE2-NEXT:    pmuludq %xmm2, %xmm4
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm10, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm14, %xmm5
 ; SSE2-NEXT:    paddq %xmm4, %xmm5
 ; SSE2-NEXT:    psllq $32, %xmm5
-; SSE2-NEXT:    pmuludq %xmm10, %xmm2
+; SSE2-NEXT:    pmuludq %xmm14, %xmm2
 ; SSE2-NEXT:    paddq %xmm5, %xmm2
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm8[2],xmm13[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm13
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm9, %xmm14
-; SSE2-NEXT:    paddq %xmm13, %xmm14
-; SSE2-NEXT:    psllq $32, %xmm14
-; SSE2-NEXT:    pmuludq %xmm9, %xmm3
-; SSE2-NEXT:    paddq %xmm14, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm3, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm8, %xmm5
+; SSE2-NEXT:    paddq %xmm4, %xmm5
+; SSE2-NEXT:    psllq $32, %xmm5
+; SSE2-NEXT:    pmuludq %xmm8, %xmm3
+; SSE2-NEXT:    paddq %xmm5, %xmm3
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: mul_v8i64_sext:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; SSE41-NEXT:    pmovsxwq %xmm3, %xmm4
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
 ; SSE41-NEXT:    pmovsxwq %xmm3, %xmm5

   %2 = sext <8 x i32> %val2 to <8 x i64>
   %3 = mul <8 x i64> %1, %2
   ret <8 x i64> %3
 }
 
 define <2 x i64> @pmuldq_square(<2 x i64> %x) {
 ; SSE2-LABEL: pmuldq_square:
 ; SSE2:       # %bb.0:
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
 ; SSE2-NEXT:    psllq $32, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,3,2,3]
 ; SSE2-NEXT:    psrad $31, %xmm0
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,3,2,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; SSE2-NEXT:    psrlq $32, %xmm0
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm1
 ; SSE2-NEXT:    paddq %xmm0, %xmm0
 ; SSE2-NEXT:    psllq $32, %xmm0

llvm/test/CodeGen/X86/pmulh.ll

   %d = lshr <4 x i32> %c, <i32 16, i32 16, i32 16, i32 16>
   ret <4 x i32> %d
 }
 
 define <4 x i32> @mulhsw_v4i16_ashr(<4 x i16> %a, <4 x i16> %b) {
 ; SSE2-LABEL: mulhsw_v4i16_ashr:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pmulhw %xmm1, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE2-NEXT:    psrad $16, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: mulhsw_v4i16_ashr:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmulhw %xmm1, %xmm0
 ; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;

   %c = mul <64 x i32> %a1, %b1
   %d = ashr <64 x i32> %c, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
   ret <64 x i32> %d
 }
 
 define <8 x i64> @mulhuw_v8i16_lshr_i64(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: mulhuw_v8i16_lshr_i64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    pxor %xmm5, %xmm5
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm5[2],xmm2[3],xmm5[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; SSE2-NEXT:    movdqa %xmm4, %xmm6
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,1,1,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,1,1,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,1,3,3]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1]
-; SSE2-NEXT:    pmuludq %xmm7, %xmm0
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm1
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; SSE2-NEXT:    movdqa %xmm3, %xmm2
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
-; SSE2-NEXT:    pmuludq %xmm6, %xmm2
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; SSE2-NEXT:    pmuludq %xmm4, %xmm3
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm4, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm3, %xmm4
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm5, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm6, %xmm3
 ; SSE2-NEXT:    psrlq $16, %xmm0
-; SSE2-NEXT:    psrlq $16, %xmm1
+; SSE2-NEXT:    psrlq $16, %xmm4
 ; SSE2-NEXT:    psrlq $16, %xmm2
 ; SSE2-NEXT:    psrlq $16, %xmm3
+; SSE2-NEXT:    movdqa %xmm4, %xmm1
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: mulhuw_v8i16_lshr_i64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmulhuw %xmm1, %xmm0
 ; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm4 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; SSE41-NEXT:    pmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero

   %c = mul <8 x i64> %a1, %b1
   %d = lshr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
   ret <8 x i64> %d
 }
 
 define <8 x i64> @mulhsw_v8i16_lshr_i64(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: mulhsw_v8i16_lshr_i64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
-; SSE2-NEXT:    psrad $16, %xmm12
-; SSE2-NEXT:    pxor %xmm8, %xmm8
-; SSE2-NEXT:    pxor %xmm14, %xmm14
-; SSE2-NEXT:    pcmpgtd %xmm12, %xmm14
-; SSE2-NEXT:    movdqa %xmm12, %xmm9
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm14[0],xmm9[1],xmm14[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm14[2],xmm12[3],xmm14[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; SSE2-NEXT:    psrad $16, %xmm6
+; SSE2-NEXT:    pxor %xmm13, %xmm13
+; SSE2-NEXT:    pxor %xmm10, %xmm10
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm10
+; SSE2-NEXT:    movdqa %xmm6, %xmm8
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm10[2],xmm6[3],xmm10[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE2-NEXT:    psrad $16, %xmm4
+; SSE2-NEXT:    pxor %xmm5, %xmm5
+; SSE2-NEXT:    pcmpgtd %xmm4, %xmm5
+; SSE2-NEXT:    movdqa %xmm4, %xmm11
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
 ; SSE2-NEXT:    psrad $16, %xmm7
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm7, %xmm4
-; SSE2-NEXT:    movdqa %xmm7, %xmm11
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
-; SSE2-NEXT:    psrad $16, %xmm5
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pcmpgtd %xmm5, %xmm0
-; SSE2-NEXT:    movdqa %xmm5, %xmm10
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE2-NEXT:    pxor %xmm12, %xmm12
+; SSE2-NEXT:    pcmpgtd %xmm7, %xmm12
+; SSE2-NEXT:    movdqa %xmm7, %xmm9
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
-; SSE2-NEXT:    movdqa %xmm1, %xmm13
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm2[0],xmm13[1],xmm2[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm7, %xmm3
-; SSE2-NEXT:    pmuludq %xmm1, %xmm7
-; SSE2-NEXT:    movdqa %xmm4, %xmm6
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm1, %xmm6
-; SSE2-NEXT:    paddq %xmm6, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm13
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm4, %xmm3
+; SSE2-NEXT:    pmuludq %xmm1, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; SSE2-NEXT:    paddq %xmm2, %xmm3
 ; SSE2-NEXT:    psllq $32, %xmm3
-; SSE2-NEXT:    paddq %xmm7, %xmm3
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
+; SSE2-NEXT:    paddq %xmm4, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,1,1,3]
 ; SSE2-NEXT:    pmuludq %xmm11, %xmm2
-; SSE2-NEXT:    pmuludq %xmm13, %xmm11
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm13, %xmm4
-; SSE2-NEXT:    paddq %xmm4, %xmm2
+; SSE2-NEXT:    pmuludq %xmm0, %xmm11
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm1
+; SSE2-NEXT:    paddq %xmm1, %xmm2
 ; SSE2-NEXT:    psllq $32, %xmm2
 ; SSE2-NEXT:    paddq %xmm11, %xmm2
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm12, %xmm1
-; SSE2-NEXT:    pmuludq %xmm5, %xmm12
-; SSE2-NEXT:    movdqa %xmm14, %xmm4
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm5, %xmm4
-; SSE2-NEXT:    paddq %xmm4, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm6, %xmm1
+; SSE2-NEXT:    pmuludq %xmm7, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm7, %xmm0
+; SSE2-NEXT:    paddq %xmm0, %xmm1
 ; SSE2-NEXT:    psllq $32, %xmm1
-; SSE2-NEXT:    paddq %xmm12, %xmm1
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm9, %xmm0
-; SSE2-NEXT:    pmuludq %xmm10, %xmm9
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm10, %xmm14
-; SSE2-NEXT:    paddq %xmm14, %xmm0
+; SSE2-NEXT:    paddq %xmm6, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm8, %xmm0
+; SSE2-NEXT:    pmuludq %xmm9, %xmm8
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm9, %xmm4
+; SSE2-NEXT:    paddq %xmm4, %xmm0
 ; SSE2-NEXT:    psllq $32, %xmm0
-; SSE2-NEXT:    paddq %xmm9, %xmm0
+; SSE2-NEXT:    paddq %xmm8, %xmm0
 ; SSE2-NEXT:    psrlq $16, %xmm0
 ; SSE2-NEXT:    psrlq $16, %xmm1
 ; SSE2-NEXT:    psrlq $16, %xmm2
 ; SSE2-NEXT:    psrlq $16, %xmm3
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: mulhsw_v8i16_lshr_i64:
 ; SSE41:       # %bb.0:

   %c = mul <8 x i64> %a1, %b1
   %d = lshr <8 x i64> %c, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
   ret <8 x i64> %d
 }
 
 define <8 x i64> @mulhsw_v8i16_ashr_i64(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: mulhsw_v8i16_ashr_i64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
-; SSE2-NEXT:    psrad $16, %xmm12
-; SSE2-NEXT:    pxor %xmm8, %xmm8
-; SSE2-NEXT:    pxor %xmm14, %xmm14
-; SSE2-NEXT:    pcmpgtd %xmm12, %xmm14
-; SSE2-NEXT:    movdqa %xmm12, %xmm9
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm14[0],xmm9[1],xmm14[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm14[2],xmm12[3],xmm14[3]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
-; SSE2-NEXT:    psrad $16, %xmm7
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm7, %xmm2
-; SSE2-NEXT:    movdqa %xmm7, %xmm11
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE2-NEXT:    psrad $16, %xmm3
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pcmpgtd %xmm3, %xmm0
-; SSE2-NEXT:    movdqa %xmm3, %xmm10
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE2-NEXT:    psrad $16, %xmm5
+; SSE2-NEXT:    pxor %xmm13, %xmm13
+; SSE2-NEXT:    pxor %xmm10, %xmm10
+; SSE2-NEXT:    pcmpgtd %xmm5, %xmm10
+; SSE2-NEXT:    movdqa %xmm5, %xmm8
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm10[0],xmm8[1],xmm10[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm10[2],xmm5[3],xmm10[3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT:    psrad $16, %xmm2
+; SSE2-NEXT:    pxor %xmm3, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm2, %xmm3
+; SSE2-NEXT:    movdqa %xmm2, %xmm11
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE2-NEXT:    psrad $16, %xmm0
+; SSE2-NEXT:    pxor %xmm12, %xmm12
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm12
+; SSE2-NEXT:    movdqa %xmm0, %xmm9
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    pxor %xmm4, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
-; SSE2-NEXT:    movdqa %xmm1, %xmm13
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1]
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; SSE2-NEXT:    movdqa %xmm4, %xmm6
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm7, %xmm6
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm13
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1]
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[2,1,3,3]
 ; SSE2-NEXT:    pmuludq %xmm1, %xmm7
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm8[2],xmm5[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm1, %xmm5
-; SSE2-NEXT:    paddq %xmm5, %xmm6
-; SSE2-NEXT:    psllq $32, %xmm6
-; SSE2-NEXT:    paddq %xmm7, %xmm6
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm11, %xmm4
-; SSE2-NEXT:    pmuludq %xmm13, %xmm11
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm13, %xmm2
-; SSE2-NEXT:    paddq %xmm2, %xmm4
+; SSE2-NEXT:    paddq %xmm7, %xmm4
 ; SSE2-NEXT:    psllq $32, %xmm4
-; SSE2-NEXT:    paddq %xmm11, %xmm4
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm12, %xmm1
-; SSE2-NEXT:    pmuludq %xmm3, %xmm12
-; SSE2-NEXT:    movdqa %xmm14, %xmm2
-; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; SSE2-NEXT:    paddq %xmm2, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm13[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm11, %xmm7
+; SSE2-NEXT:    pmuludq %xmm6, %xmm11
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm6, %xmm1
+; SSE2-NEXT:    paddq %xmm1, %xmm7
+; SSE2-NEXT:    psllq $32, %xmm7
+; SSE2-NEXT:    paddq %xmm11, %xmm7
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm5, %xmm1
+; SSE2-NEXT:    pmuludq %xmm0, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[2,1,3,3]
+; SSE2-NEXT:    pmuludq %xmm0, %xmm2
 ; SSE2-NEXT:    paddq %xmm2, %xmm1
 ; SSE2-NEXT:    psllq $32, %xmm1
-; SSE2-NEXT:    paddq %xmm12, %xmm1
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm9, %xmm0
-; SSE2-NEXT:    pmuludq %xmm10, %xmm9
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
-; SSE2-NEXT:    pmuludq %xmm10, %xmm14
-; SSE2-NEXT:    paddq %xmm14, %xmm0
+; SSE2-NEXT:    paddq %xmm5, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm8, %xmm0
+; SSE2-NEXT:    pmuludq %xmm9, %xmm8
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,1,1,3]
+; SSE2-NEXT:    pmuludq %xmm9, %xmm2
+; SSE2-NEXT:    paddq %xmm2, %xmm0
 ; SSE2-NEXT:    psllq $32, %xmm0
-; SSE2-NEXT:    paddq %xmm9, %xmm0
+; SSE2-NEXT:    paddq %xmm8, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    psrad $16, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; SSE2-NEXT:    psrlq $16, %xmm0
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    psrad $16, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; SSE2-NEXT:    psrlq $16, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE2-NEXT:    movdqa %xmm4, %xmm2
+; SSE2-NEXT:    movdqa %xmm7, %xmm2
 ; SSE2-NEXT:    psrad $16, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,3,2,3]
-; SSE2-NEXT:    psrlq $16, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
+; SSE2-NEXT:    psrlq $16, %xmm7
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE2-NEXT:    movdqa %xmm6, %xmm3
+; SSE2-NEXT:    movdqa %xmm4, %xmm3
 ; SSE2-NEXT:    psrad $16, %xmm3
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,3,2,3]
-; SSE2-NEXT:    psrlq $16, %xmm6
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[1,3,2,3]
+; SSE2-NEXT:    psrlq $16, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: mulhsw_v8i16_ashr_i64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmulhw %xmm1, %xmm0
 ; SSE41-NEXT:    pmovsxwq %xmm0, %xmm4
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
 ; SSE41-NEXT:    pmovsxwq %xmm1, %xmm1

llvm/test/CodeGen/X86/pr29112.ll

 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm1 = xmm11[0,1],xmm2[1],xmm11[3]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm13 = xmm1[0,1,2],xmm3[1]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm6 = xmm4[0,1,2],xmm3[1]
 ; CHECK-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    vextractf32x4 $2, %zmm3, %xmm4
 ; CHECK-NEXT:    vblendps {{.*#+}} xmm4 = xmm1[0,1,2],xmm4[3]
 ; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[3,3,3,3]
 ; CHECK-NEXT:    vunpcklps {{.*#+}} xmm5 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; CHECK-NEXT:    vinsertps {{.*#+}} xmm5 = xmm5[0,1],xmm2[1],xmm5[3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm5 = xmm5[0,1],xmm2[1,3]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm5 = xmm5[0,1,2],xmm3[1]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm0 = xmm7[0,1],xmm2[1],xmm7[3]
 ; CHECK-NEXT:    vblendps {{.*#+}} xmm7 = xmm0[0,1,2],xmm3[3]
 ; CHECK-NEXT:    vblendps {{.*#+}} xmm12 = xmm1[0,1,2],xmm3[3]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[1]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[1]
 ; CHECK-NEXT:    vaddps %xmm1, %xmm0, %xmm8
-; CHECK-NEXT:    vinsertps {{.*#+}} xmm2 = xmm11[0,1],xmm2[3],xmm11[3]
+; CHECK-NEXT:    vshufps {{.*#+}} xmm2 = xmm11[0,1],xmm2[3,3]
 ; CHECK-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[2]
 ; CHECK-NEXT:    vaddps %xmm2, %xmm14, %xmm2
 ; CHECK-NEXT:    vmovaps %xmm13, %xmm1
 ; CHECK-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    vaddps %xmm10, %xmm13, %xmm10
 ; CHECK-NEXT:    vaddps %xmm13, %xmm13, %xmm3
 ; CHECK-NEXT:    vaddps %xmm12, %xmm14, %xmm0
 ; CHECK-NEXT:    vaddps %xmm0, %xmm8, %xmm0

llvm/test/CodeGen/X86/promote-cmp.ll

 ; SSE2-NEXT:    pxor %xmm0, %xmm4
 ; SSE2-NEXT:    movdqa %xmm4, %xmm7
 ; SSE2-NEXT:    pcmpgtd %xmm5, %xmm7
 ; SSE2-NEXT:    pcmpeqd %xmm5, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
 ; SSE2-NEXT:    pand %xmm7, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
 ; SSE2-NEXT:    por %xmm4, %xmm5
-; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm6[0,2]
-; SSE2-NEXT:    movaps {{.*#+}} xmm4 = <1,1,u,0>
-; SSE2-NEXT:    xorps %xmm5, %xmm4
-; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1,3,3]
-; SSE2-NEXT:    psllq $63, %xmm5
-; SSE2-NEXT:    psrad $31, %xmm5
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,2,2,3]
+; SSE2-NEXT:    pxor {{.*}}(%rip), %xmm4
+; SSE2-NEXT:    psllq $63, %xmm6
+; SSE2-NEXT:    psrad $31, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3]
 ; SSE2-NEXT:    pand %xmm5, %xmm1
 ; SSE2-NEXT:    pandn %xmm3, %xmm5
 ; SSE2-NEXT:    por %xmm5, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,1,1,3]
 ; SSE2-NEXT:    psllq $63, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
 ; SSE2-NEXT:    pand %xmm3, %xmm0

llvm/test/CodeGen/X86/psubus.ll

 ; AVX1-NEXT:    vpmaxud %xmm4, %xmm2, %xmm5
 ; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm2, %xmm2
 ; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
 ; AVX1-NEXT:    vpxor %xmm5, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmaxud %xmm1, %xmm3, %xmm6
 ; AVX1-NEXT:    vpcmpeqd %xmm6, %xmm3, %xmm3
 ; AVX1-NEXT:    vpxor %xmm5, %xmm3, %xmm3
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm4, %xmm4
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
 ; AVX1-NEXT:    vpsubw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpandn %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;

 ; AVX1-NEXT:    vpminud %xmm4, %xmm2, %xmm5
 ; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm2, %xmm2
 ; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
 ; AVX1-NEXT:    vpxor %xmm5, %xmm2, %xmm2
 ; AVX1-NEXT:    vpminud %xmm1, %xmm3, %xmm6
 ; AVX1-NEXT:    vpcmpeqd %xmm6, %xmm3, %xmm3
 ; AVX1-NEXT:    vpxor %xmm5, %xmm3, %xmm3
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm4, %xmm4
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
 ; AVX1-NEXT:    vpsubw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;

 ; AVX1-NEXT:    vpmaxud %xmm2, %xmm4, %xmm2
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm4, %xmm2
 ; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
 ; AVX1-NEXT:    vpxor %xmm5, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmaxud %xmm3, %xmm1, %xmm3
 ; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm1, %xmm3
 ; AVX1-NEXT:    vpxor %xmm5, %xmm3, %xmm3
 ; AVX1-NEXT:    vpackssdw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm4, %xmm4
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
 ; AVX1-NEXT:    vpsubw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand %xmm0, %xmm2, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;

llvm/test/CodeGen/X86/shrink_vmul.ll

 ; X86-SSE-NEXT:    movd %edx, %xmm0
 ; X86-SSE-NEXT:    movzwl (%eax,%ecx), %eax
 ; X86-SSE-NEXT:    movd %eax, %xmm1
 ; X86-SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; X86-SSE-NEXT:    psraw $8, %xmm0
 ; X86-SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; X86-SSE-NEXT:    psraw $8, %xmm1
 ; X86-SSE-NEXT:    pmullw %xmm0, %xmm1
-; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,0,2,1,4,5,6,7]
 ; X86-SSE-NEXT:    psrad $16, %xmm0
 ; X86-SSE-NEXT:    movq %xmm0, (%esi,%ecx,4)
 ; X86-SSE-NEXT:    popl %esi
 ; X86-SSE-NEXT:    retl
 ;
 ; X86-AVX-LABEL: mul_2xi8_sext:
 ; X86-AVX:       # %bb.0: # %entry
 ; X86-AVX-NEXT:    pushl %esi

 ; X64-SSE-NEXT:    movd %ecx, %xmm0
 ; X64-SSE-NEXT:    movzwl (%rsi,%rdx), %ecx
 ; X64-SSE-NEXT:    movd %ecx, %xmm1
 ; X64-SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; X64-SSE-NEXT:    psraw $8, %xmm0
 ; X64-SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; X64-SSE-NEXT:    psraw $8, %xmm1
 ; X64-SSE-NEXT:    pmullw %xmm0, %xmm1
-; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X64-SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,0,2,1,4,5,6,7]
 ; X64-SSE-NEXT:    psrad $16, %xmm0
 ; X64-SSE-NEXT:    movq %xmm0, (%rax,%rdx,4)
 ; X64-SSE-NEXT:    retq
 ;
 ; X64-AVX-LABEL: mul_2xi8_sext:
 ; X64-AVX:       # %bb.0: # %entry
 ; X64-AVX-NEXT:    movq {{.*}}(%rip), %rax
 ; X64-AVX-NEXT:    movzwl (%rdi,%rdx), %ecx

 ; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-SSE-NEXT:    movl c, %edx
 ; X86-SSE-NEXT:    movzwl (%ecx,%eax), %ecx
 ; X86-SSE-NEXT:    movd %ecx, %xmm0
 ; X86-SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; X86-SSE-NEXT:    psraw $8, %xmm0
 ; X86-SSE-NEXT:    pmullw {{\.LCPI.*}}, %xmm0
-; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X86-SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; X86-SSE-NEXT:    psrad $16, %xmm0
 ; X86-SSE-NEXT:    movq %xmm0, (%edx,%eax,4)
 ; X86-SSE-NEXT:    retl
 ;
 ; X86-AVX-LABEL: mul_2xi8_varconst2:
 ; X86-AVX:       # %bb.0: # %entry
 ; X86-AVX-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-AVX-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-AVX-NEXT:    movl c, %edx
 ; X86-AVX-NEXT:    movzwl (%ecx,%eax), %ecx
 ; X86-AVX-NEXT:    vmovd %ecx, %xmm0
 ; X86-AVX-NEXT:    vpmovsxbd %xmm0, %xmm0
 ; X86-AVX-NEXT:    vpmulld {{\.LCPI.*}}, %xmm0, %xmm0
 ; X86-AVX-NEXT:    vmovq %xmm0, (%edx,%eax,4)
 ; X86-AVX-NEXT:    retl
 ;
 ; X64-SSE-LABEL: mul_2xi8_varconst2:
 ; X64-SSE:       # %bb.0: # %entry
 ; X64-SSE-NEXT:    movq {{.*}}(%rip), %rax
 ; X64-SSE-NEXT:    movzwl (%rdi,%rsi), %ecx
 ; X64-SSE-NEXT:    movd %ecx, %xmm0
 ; X64-SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; X64-SSE-NEXT:    psraw $8, %xmm0
 ; X64-SSE-NEXT:    pmullw {{.*}}(%rip), %xmm0
-; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; X64-SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; X64-SSE-NEXT:    psrad $16, %xmm0
 ; X64-SSE-NEXT:    movq %xmm0, (%rax,%rsi,4)
 ; X64-SSE-NEXT:    retq
 ;
 ; X64-AVX-LABEL: mul_2xi8_varconst2:
 ; X64-AVX:       # %bb.0: # %entry
 ; X64-AVX-NEXT:    movq {{.*}}(%rip), %rax
 ; X64-AVX-NEXT:    movzwl (%rdi,%rsi), %ecx

llvm/test/CodeGen/X86/shuffle-of-insert.ll

; AVX-NEXT:    retq
   %shuf = shufflevector <4 x i32> %ins, <4 x i32> %v2, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
   ret <4 x i32> %shuf
 }
 
 define <4 x i32> @ins_elt_1(i32 %x, <4 x i32> %v1, <4 x i32> %v2) {
 ; SSE2-LABEL: ins_elt_1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movd %edi, %xmm0
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    retq
 ;
 ; SSE4-LABEL: ins_elt_1:
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    movdqa %xmm1, %xmm0
 ; SSE4-NEXT:    pinsrd $1, %edi, %xmm0
 ; SSE4-NEXT:    retq
; AVX-NEXT:    retq
   ret <4 x i32> %shuf
 }
 
 define <4 x i32> @ins_elt_3_commute(i32 %x, <4 x i32> %v1, <4 x i32> %v2) {
 ; SSE2-LABEL: ins_elt_3_commute:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    movd %edi, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE4-LABEL: ins_elt_3_commute:
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    movdqa %xmm1, %xmm0
 ; SSE4-NEXT:    pinsrd $3, %edi, %xmm0
 ; SSE4-NEXT:    retq
; AVX-NEXT:    retq
   ret <4 x i32> %shuf
 }
 
 define <4 x i32> @ins_elt_2_to_3(i32 %x, <4 x i32> %v1, <4 x i32> %v2) {
 ; SSE2-LABEL: ins_elt_2_to_3:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    movd %edi, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE4-LABEL: ins_elt_2_to_3:
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    movdqa %xmm1, %xmm0
 ; SSE4-NEXT:    pinsrd $3, %edi, %xmm0
 ; SSE4-NEXT:    retq
 ;
 ; AVX-LABEL: ins_elt_2_to_3:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpinsrd $3, %edi, %xmm1, %xmm0
 ; AVX-NEXT:    retq
   %ins = insertelement <4 x i32> %v1, i32 %x, i32 2
   %shuf = shufflevector <4 x i32> %v2, <4 x i32> %ins, <4 x i32> <i32 0, i32 1, i32 2, i32 6>
   ret <4 x i32> %shuf
 }
 
 define <4 x i32> @ins_elt_3_to_1(i32 %x, <4 x i32> %v1, <4 x i32> %v2) {
 ; SSE2-LABEL: ins_elt_3_to_1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movd %edi, %xmm0
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    retq
 ;
 ; SSE4-LABEL: ins_elt_3_to_1:
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    movdqa %xmm1, %xmm0
 ; SSE4-NEXT:    pinsrd $1, %edi, %xmm0
 ; SSE4-NEXT:    retq

llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE42
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 --check-prefix=AVX2-SLOW
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 --check-prefix=AVX2-FAST
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VL
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BW
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BWVL
 
 define void @shuffle_v16i8_to_v8i8_1(<16 x i8>* %L, <8 x i8>* %S) nounwind {
 ; SSE2-LABEL: shuffle_v16i8_to_v8i8_1:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa (%rdi), %xmm0
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE2-NEXT:    psrlw $8, %xmm0
 ; SSE2-NEXT:    packuswb %xmm0, %xmm0
 ; SSE2-NEXT:    movq %xmm0, (%rsi)
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: shuffle_v16i8_to_v8i8_1:
 ; SSE42:       # %bb.0:
 ; SSE42-NEXT:    movdqa (%rdi), %xmm0
 ; SSE42-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u]

 }
 
 define void @shuffle_v16i8_to_v2i8_7(<16 x i8>* %L, <2 x i8>* %S) nounwind {
 ; SSE2-LABEL: shuffle_v16i8_to_v2i8_7:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa (%rdi), %xmm0
 ; SSE2-NEXT:    pxor %xmm1, %xmm1
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
 ; SSE2-NEXT:    packuswb %xmm0, %xmm0
 ; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    movw %ax, (%rsi)
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: shuffle_v16i8_to_v2i8_7:
 ; SSE42:       # %bb.0:
 ; SSE42-NEXT:    movdqa (%rdi), %xmm0

llvm/test/CodeGen/X86/shuffle-vs-trunc-256.ll

   ret void
 }
 
 define void @trunc_v8i32_to_v8i16(<16 x i16>* %L, <8 x i16>* %S) nounwind {
 ; AVX1-LABEL: trunc_v8i32_to_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX1-NEXT:    vmovdqa 16(%rdi), %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vmovdqa %xmm0, (%rsi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_v8i32_to_v8i16:
 ; AVX2:       # %bb.0:

 ; AVX512BWVL-NEXT:    vpmovdb %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX512BWVL-NEXT:    vmovq %xmm0, (%rsi)
 ; AVX512BWVL-NEXT:    retq
 ;
 ; AVX512VBMIVL-LABEL: shuffle_v32i8_to_v8i8:
 ; AVX512VBMIVL:       # %bb.0:
 ; AVX512VBMIVL-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512VBMIVL-NEXT:    vmovdqa 16(%rdi), %xmm1
-; AVX512VBMIVL-NEXT:    vpmovdb %xmm1, %xmm1
-; AVX512VBMIVL-NEXT:    vpmovdb %xmm0, %xmm0
-; AVX512VBMIVL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX512VBMIVL-NEXT:    vmovq %xmm0, (%rsi)
+; AVX512VBMIVL-NEXT:    vpbroadcastq {{.*#+}} xmm1 = [2024390091656922112,2024390091656922112]
+; AVX512VBMIVL-NEXT:    vpermi2b 16(%rdi), %xmm0, %xmm1
+; AVX512VBMIVL-NEXT:    vmovq %xmm1, (%rsi)
 ; AVX512VBMIVL-NEXT:    retq
   %vec = load <32 x i8>, <32 x i8>* %L
   %strided.vec = shufflevector <32 x i8> %vec, <32 x i8> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
   store <8 x i8> %strided.vec, <8 x i8>* %S
   ret void
 }
 
 define void @trunc_v8i32_to_v8i8(<32 x i8>* %L, <8 x i8>* %S) nounwind {

 ; AVX512F-NEXT:    vpackusdw %xmm0, %xmm0, %xmm0
 ; AVX512F-NEXT:    vmovq %xmm0, (%rsi)
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_to_v4i16:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa (%rdi), %xmm0
 ; AVX512VL-NEXT:    vmovdqa 16(%rdi), %xmm1
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,8,9,8,9,10,11,8,9,10,11,12,13,14,15]
-; AVX512VL-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpmovqw %xmm1, %xmm1
+; AVX512VL-NEXT:    vpmovqw %xmm0, %xmm0
 ; AVX512VL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX512VL-NEXT:    vmovq %xmm0, (%rsi)
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: shuffle_v16i16_to_v4i16:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpblendw {{.*#+}} xmm1 = mem[0],xmm0[1,2,3],mem[4],xmm0[5,6,7]

 ; AVX512BWVL-NEXT:    vpmovqb %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX512BWVL-NEXT:    vmovd %xmm0, (%rsi)
 ; AVX512BWVL-NEXT:    retq
 ;
 ; AVX512VBMIVL-LABEL: shuffle_v32i8_to_v4i8:
 ; AVX512VBMIVL:       # %bb.0:
 ; AVX512VBMIVL-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512VBMIVL-NEXT:    vmovdqa 16(%rdi), %xmm1
-; AVX512VBMIVL-NEXT:    vpmovqb %xmm1, %xmm1
-; AVX512VBMIVL-NEXT:    vpmovqb %xmm0, %xmm0
-; AVX512VBMIVL-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX512VBMIVL-NEXT:    vmovd %xmm0, (%rsi)
+; AVX512VBMIVL-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [403703808,403703808,403703808,403703808]
+; AVX512VBMIVL-NEXT:    vpermi2b 16(%rdi), %xmm0, %xmm1
+; AVX512VBMIVL-NEXT:    vmovd %xmm1, (%rsi)
 ; AVX512VBMIVL-NEXT:    retq
   %vec = load <32 x i8>, <32 x i8>* %L
   %strided.vec = shufflevector <32 x i8> %vec, <32 x i8> undef, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
   store <4 x i8> %strided.vec, <4 x i8>* %S
   ret void
 }
 
 define void @trunc_v4i64_to_v4i8(<32 x i8>* %L, <4 x i8>* %S) nounwind {

llvm/test/CodeGen/X86/shuffle-vs-trunc-512.ll

 ; AVX512F-NEXT:    vpblendw {{.*#+}} xmm0 = mem[0],xmm0[1,2,3],mem[4],xmm0[5,6,7]
 ; AVX512F-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX512F-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX512F-NEXT:    vmovdqa %xmm0, (%rsi)
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v32i16_to_v8i16:
 ; AVX512VL:       # %bb.0:
-; AVX512VL-NEXT:    vmovdqa (%rdi), %xmm0
-; AVX512VL-NEXT:    vmovdqa 16(%rdi), %xmm1
+; AVX512VL-NEXT:    vmovdqa 48(%rdi), %xmm0
+; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,8,9,8,9,10,11,12,13,14,15]
+; AVX512VL-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX512VL-NEXT:    vmovdqa 32(%rdi), %xmm2
-; AVX512VL-NEXT:    vmovdqa 48(%rdi), %xmm3
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,2,3,0,1,8,9,8,9,10,11,12,13,14,15]
-; AVX512VL-NEXT:    vpshufb %xmm4, %xmm3, %xmm3
-; AVX512VL-NEXT:    vpshufb %xmm4, %xmm2, %xmm2
-; AVX512VL-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,8,9,8,9,10,11,8,9,10,11,12,13,14,15]
-; AVX512VL-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX512VL-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
-; AVX512VL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX512VL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
+; AVX512VL-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
+; AVX512VL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX512VL-NEXT:    vmovdqa (%rdi), %ymm1
+; AVX512VL-NEXT:    vpmovqw %ymm1, %xmm1
+; AVX512VL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX512VL-NEXT:    vmovdqa %xmm0, (%rsi)
+; AVX512VL-NEXT:    vzeroupper
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: shuffle_v32i16_to_v8i16:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpblendw {{.*#+}} xmm1 = mem[0],xmm0[1,2,3],mem[4],xmm0[5,6,7]
 ; AVX512BW-NEXT:    vpblendw {{.*#+}} xmm2 = mem[0],xmm0[1,2,3],mem[4],xmm0[5,6,7]
 ; AVX512BW-NEXT:    vpackusdw %xmm1, %xmm2, %xmm1

 ; AVX512BWVL-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; AVX512BWVL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX512BWVL-NEXT:    vzeroupper
 ; AVX512BWVL-NEXT:    retq
 ;
 ; AVX512VBMI-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
 ; AVX512VBMI:       # %bb.0:
-; AVX512VBMI-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
-; AVX512VBMI-NEXT:    vpermb %zmm0, %zmm1, %zmm0
-; AVX512VBMI-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VBMI-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [5569058560453190945,5569058560453190945,5569058560453190945,5569058560453190945,5569058560453190945,5569058560453190945,5569058560453190945,5569058560453190945]
+; AVX512VBMI-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
+; AVX512VBMI-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX512VBMI-NEXT:    vpermi2b %zmm2, %zmm0, %zmm1
+; AVX512VBMI-NEXT:    vpbroadcastq {{.*#+}} zmm2 = [6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281]
+; AVX512VBMI-NEXT:    vpermi2b %zmm0, %zmm0, %zmm2
+; AVX512VBMI-NEXT:    vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm1[2,3]
 ; AVX512VBMI-NEXT:    vzeroupper
 ; AVX512VBMI-NEXT:    retq
 ;
 ; AVX512VBMIVL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:
 ; AVX512VBMIVL:       # %bb.0:
 ; AVX512VBMIVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,61]
 ; AVX512VBMIVL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
 ; AVX512VBMIVL-NEXT:    vpermt2b %ymm2, %ymm1, %ymm0

 ; AVX512BWVL-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,1,5,9,13,u,u,u,u,u,u,u,u]
 ; AVX512BWVL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; AVX512BWVL-NEXT:    vpblendd {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
 ; AVX512BWVL-NEXT:    vzeroupper
 ; AVX512BWVL-NEXT:    retq
 ;
 ; AVX512VBMI-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:
 ; AVX512VBMI:       # %bb.0:
-; AVX512VBMI-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,62]
-; AVX512VBMI-NEXT:    vpermb %zmm0, %zmm1, %zmm0
-; AVX512VBMI-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VBMI-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [5641116154491118881,5641116154491118881,5641116154491118881,5641116154491118881,5641116154491118881,5641116154491118881,5641116154491118881,5641116154491118881]
+; AVX512VBMI-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
+; AVX512VBMI-NEXT:    vextracti128 $1, %ymm2, %xmm2
+; AVX512VBMI-NEXT:    vpermi2b %zmm2, %zmm0, %zmm1
+; AVX512VBMI-NEXT:    vpbroadcastq {{.*#+}} zmm2 = [6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281,6726501325053953281]
+; AVX512VBMI-NEXT:    vpermi2b %zmm0, %zmm0, %zmm2
+; AVX512VBMI-NEXT:    vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm1[2,3]
 ; AVX512VBMI-NEXT:    vzeroupper
 ; AVX512VBMI-NEXT:    retq
 ;
 ; AVX512VBMIVL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:
 ; AVX512VBMIVL:       # %bb.0:
 ; AVX512VBMIVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,62]
 ; AVX512VBMIVL-NEXT:    vextracti64x4 $1, %zmm0, %ymm2
 ; AVX512VBMIVL-NEXT:    vpermt2b %ymm2, %ymm1, %ymm0
 ; AVX512VBMIVL-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; AVX512VBMIVL-NEXT:    vzeroupper
 ; AVX512VBMIVL-NEXT:    retq
   %res = shufflevector <64 x i8> %x, <64 x i8> %x, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 62>
   ret <16 x i8> %res
 }
 
 define <4 x double> @PR34175(<32 x i16>* %p) {
 ; AVX512F-LABEL: PR34175:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vmovdqu (%rdi), %xmm0
 ; AVX512F-NEXT:    vmovdqu 32(%rdi), %xmm1
-; AVX512F-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX512F-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX512F-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX512F-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX512F-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX512F-NEXT:    vcvtdq2pd %xmm0, %ymm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: PR34175:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqu (%rdi), %xmm0
 ; AVX512VL-NEXT:    vmovdqu 32(%rdi), %xmm1
 ; AVX512VL-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
 ; AVX512VL-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX512VL-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX512VL-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX512VL-NEXT:    vcvtdq2pd %xmm0, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; AVX512BW-LABEL: PR34175:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vmovdqu (%rdi), %xmm0
 ; AVX512BW-NEXT:    vmovdqu 32(%rdi), %xmm1
-; AVX512BW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX512BW-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX512BW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX512BW-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX512BW-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX512BW-NEXT:    vcvtdq2pd %xmm0, %ymm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512BWVL-LABEL: PR34175:
 ; AVX512BWVL:       # %bb.0:
 ; AVX512BWVL-NEXT:    vmovdqu (%rdi), %ymm0
 ; AVX512BWVL-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6755468161056768,6755468161056768,6755468161056768,6755468161056768]
 ; AVX512BWVL-NEXT:    vpermi2w 32(%rdi), %ymm0, %ymm1
 ; AVX512BWVL-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
 ; AVX512BWVL-NEXT:    vcvtdq2pd %xmm0, %ymm0
 ; AVX512BWVL-NEXT:    retq
 ;
 ; AVX512VBMI-LABEL: PR34175:
 ; AVX512VBMI:       # %bb.0:
 ; AVX512VBMI-NEXT:    vmovdqu (%rdi), %xmm0
 ; AVX512VBMI-NEXT:    vmovdqu 32(%rdi), %xmm1
-; AVX512VBMI-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
 ; AVX512VBMI-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX512VBMI-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX512VBMI-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX512VBMI-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX512VBMI-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; AVX512VBMI-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX512VBMI-NEXT:    vcvtdq2pd %xmm0, %ymm0
 ; AVX512VBMI-NEXT:    retq
 ;
 ; AVX512VBMIVL-LABEL: PR34175:
 ; AVX512VBMIVL:       # %bb.0:
 ; AVX512VBMIVL-NEXT:    vmovdqu (%rdi), %ymm0
 ; AVX512VBMIVL-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6755468161056768,6755468161056768,6755468161056768,6755468161056768]

llvm/test/CodeGen/X86/srem-seteq-vec-nonsplat.ll

   %ret = zext <4 x i1> %cmp to <4 x i32>
   ret <4 x i32> %ret
 }
 
 ; One power-of-two divisor in odd+even divisor
 define <4 x i32> @test_srem_odd_even_poweroftwo(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_srem_odd_even_poweroftwo:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,0]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1717986919,2454267027,2147483649,1374389535]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1717986919,2454267027,2147483649,1374389535]
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pmuludq %xmm5, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pxor %xmm4, %xmm4
-; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm4
-; CHECK-SSE2-NEXT:    pand %xmm5, %xmm4
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [0,4294967295,4294967295,0]
-; CHECK-SSE2-NEXT:    pand %xmm0, %xmm5
-; CHECK-SSE2-NEXT:    paddd %xmm4, %xmm5
-; CHECK-SSE2-NEXT:    psubd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    pxor %xmm5, %xmm5
+; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
+; CHECK-SSE2-NEXT:    pand %xmm2, %xmm5
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,4294967295,4294967295,0]
+; CHECK-SSE2-NEXT:    pand %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    paddd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,1,0]
+; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
 ; CHECK-SSE2-NEXT:    paddd %xmm3, %xmm2
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
 ; CHECK-SSE2-NEXT:    psrad $5, %xmm3
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    psrad $3, %xmm4
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm5
 ; CHECK-SSE2-NEXT:    psrad $1, %xmm5
 ; CHECK-SSE2-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]

   %ret = zext <4 x i1> %cmp to <4 x i32>
   ret <4 x i32> %ret
 }
 
 ; One one divisor in odd+even divisor
 define <4 x i32> @test_srem_odd_even_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_srem_odd_even_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,0]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1717986919,2454267027,0,1374389535]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1717986919,2454267027,0,1374389535]
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pmuludq %xmm5, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pxor %xmm4, %xmm4
-; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm4
-; CHECK-SSE2-NEXT:    pand %xmm5, %xmm4
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [0,4294967295,0,0]
-; CHECK-SSE2-NEXT:    pand %xmm0, %xmm5
-; CHECK-SSE2-NEXT:    paddd %xmm4, %xmm5
-; CHECK-SSE2-NEXT:    psubd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    pxor %xmm5, %xmm5
+; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
+; CHECK-SSE2-NEXT:    pand %xmm2, %xmm5
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,4294967295,0,0]
+; CHECK-SSE2-NEXT:    pand %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    paddd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,1,0]
+; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
 ; CHECK-SSE2-NEXT:    paddd %xmm3, %xmm2
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
 ; CHECK-SSE2-NEXT:    psrad $5, %xmm3
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
 ; CHECK-SSE2-NEXT:    psrad $3, %xmm3
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm5

 }
 
 ;------------------------------------------------------------------------------;
 
 ; One power-of-two divisor divisor and one divisor in odd divisor
 define <4 x i32> @test_srem_odd_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_srem_odd_poweroftwo_and_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,0]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1717986919,2147483649,0,1717986919]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1717986919,2147483649,0,1717986919]
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pmuludq %xmm5, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pxor %xmm4, %xmm4
-; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm4
-; CHECK-SSE2-NEXT:    pand %xmm5, %xmm4
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [0,4294967295,0,0]
-; CHECK-SSE2-NEXT:    pand %xmm0, %xmm5
-; CHECK-SSE2-NEXT:    paddd %xmm4, %xmm5
-; CHECK-SSE2-NEXT:    psubd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    pxor %xmm5, %xmm5
+; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
+; CHECK-SSE2-NEXT:    pand %xmm2, %xmm5
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,4294967295,0,0]
+; CHECK-SSE2-NEXT:    pand %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    paddd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,1,0]
+; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
 ; CHECK-SSE2-NEXT:    paddd %xmm3, %xmm2
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
 ; CHECK-SSE2-NEXT:    psrad $1, %xmm3
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm5
 ; CHECK-SSE2-NEXT:    psrad $3, %xmm5
 ; CHECK-SSE2-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]

   %ret = zext <4 x i1> %cmp to <4 x i32>
   ret <4 x i32> %ret
 }
 
 ; One power-of-two divisor divisor and one divisor in odd+even divisor
 define <4 x i32> @test_srem_odd_even_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_srem_odd_even_poweroftwo_and_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,0]
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [1717986919,2147483649,0,1374389535]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1717986919,2147483649,0,1374389535]
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pmuludq %xmm5, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
-; CHECK-SSE2-NEXT:    pxor %xmm4, %xmm4
-; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm4
-; CHECK-SSE2-NEXT:    pand %xmm5, %xmm4
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [0,4294967295,0,0]
-; CHECK-SSE2-NEXT:    pand %xmm0, %xmm5
-; CHECK-SSE2-NEXT:    paddd %xmm4, %xmm5
-; CHECK-SSE2-NEXT:    psubd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    pxor %xmm5, %xmm5
+; CHECK-SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
+; CHECK-SSE2-NEXT:    pand %xmm2, %xmm5
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,4294967295,0,0]
+; CHECK-SSE2-NEXT:    pand %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    paddd %xmm5, %xmm2
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,1,0]
+; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
 ; CHECK-SSE2-NEXT:    paddd %xmm3, %xmm2
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
 ; CHECK-SSE2-NEXT:    psrad $5, %xmm3
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm3
 ; CHECK-SSE2-NEXT:    psrad $3, %xmm3
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm5

llvm/test/CodeGen/X86/test-shrink-bug.ll

 ; CHECK-X64-LABEL: fail:
 ; CHECK-X64:       # %bb.0:
 ; CHECK-X64-NEXT:    pushq %rax
 ; CHECK-X64-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-X64-NEXT:    testl $263, %edi # imm = 0x107
 ; CHECK-X64-NEXT:    je .LBB1_3
 ; CHECK-X64-NEXT:  # %bb.1:
 ; CHECK-X64-NEXT:    pcmpeqb {{.*}}(%rip), %xmm0
-; CHECK-X64-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; CHECK-X64-NEXT:    pextrw $1, %xmm0, %eax
+; CHECK-X64-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8]
+; CHECK-X64-NEXT:    pextrw $4, %xmm0, %eax
 ; CHECK-X64-NEXT:    testb $1, %al
 ; CHECK-X64-NEXT:    jne .LBB1_3
 ; CHECK-X64-NEXT:  # %bb.2: # %no
 ; CHECK-X64-NEXT:    callq bar
 ; CHECK-X64-NEXT:  .LBB1_3: # %yes
 ; CHECK-X64-NEXT:    popq %rax
 ; CHECK-X64-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-X64-NEXT:    retq

llvm/test/CodeGen/X86/trunc-subvector.ll

 ; SSE2-LABEL: test5:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    retq
 ;
 ; AVX2-LABEL: test5:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpmovzxdq {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
-; AVX2-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [17179869187,17179869187,17179869187,17179869187]
+; AVX2-NEXT:    vpermps %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test5:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vpmovsxdq %ymm0, %zmm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1

 ; SSE2-LABEL: test10:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    retq
 ;
 ; AVX2-LABEL: test10:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpmovzxdq {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
-; AVX2-NEXT:    vpermd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vbroadcastsd {{.*#+}} ymm1 = [17179869187,17179869187,17179869187,17179869187]
+; AVX2-NEXT:    vpermps %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: test10:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vpmovzxdq {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1

llvm/test/CodeGen/X86/udiv_fix.ll

   %tmp = call i16 @llvm.udiv.fix.i16(i16 %x, i16 %y, i32 16)
   ret i16 %tmp
 }
 
 define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X64-LABEL: vec:
 ; X64:       # %bb.0:
 ; X64-NEXT:    pxor %xmm2, %xmm2
-; X64-NEXT:    movdqa %xmm1, %xmm4
+; X64-NEXT:    movdqa %xmm1, %xmm3
+; X64-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; X64-NEXT:    movq %xmm3, %rcx
+; X64-NEXT:    movdqa %xmm0, %xmm4
 ; X64-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; X64-NEXT:    movq %xmm4, %rcx
-; X64-NEXT:    movdqa %xmm0, %xmm5
-; X64-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; X64-NEXT:    psllq $31, %xmm5
-; X64-NEXT:    movq %xmm5, %rax
+; X64-NEXT:    psllq $31, %xmm4
+; X64-NEXT:    movq %xmm4, %rax
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm3
 ; X64-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; X64-NEXT:    movq %xmm4, %rcx
-; X64-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[2,3,2,3]
 ; X64-NEXT:    movq %xmm4, %rax
+; X64-NEXT:    movdqa %xmm1, %xmm4
+; X64-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    movq %xmm4, %rcx
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm4
 ; X64-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0]
-; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X64-NEXT:    movq %xmm1, %rcx
 ; X64-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X64-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; X64-NEXT:    movq %xmm2, %rcx
 ; X64-NEXT:    psllq $31, %xmm0
 ; X64-NEXT:    movq %xmm0, %rax
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm2
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X64-NEXT:    movq %xmm1, %rcx
 ; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; X64-NEXT:    movq %xmm0, %rax
+; X64-NEXT:    psrlq $32, %xmm1
+; X64-NEXT:    movq %xmm1, %rcx
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm0
 ; X64-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm0[0]
 ; X64-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
 ; X64-NEXT:    movaps %xmm2, %xmm0
 ; X64-NEXT:    retq
 ;

llvm/test/CodeGen/X86/udiv_fix_sat.ll

 ; X86-NEXT:    retl
   %tmp = call i16 @llvm.udiv.fix.sat.i16(i16 %x, i16 %y, i32 16)
   ret i16 %tmp
 }
 
 define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X64-LABEL: vec:
 ; X64:       # %bb.0:
-; X64-NEXT:    pxor %xmm8, %xmm8
+; X64-NEXT:    pxor %xmm3, %xmm3
 ; X64-NEXT:    movdqa %xmm1, %xmm2
-; X64-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; X64-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
 ; X64-NEXT:    movq %xmm2, %rcx
-; X64-NEXT:    movdqa %xmm0, %xmm4
-; X64-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; X64-NEXT:    paddq %xmm4, %xmm4
-; X64-NEXT:    psllq $31, %xmm4
-; X64-NEXT:    movq %xmm4, %rax
+; X64-NEXT:    movdqa %xmm0, %xmm2
+; X64-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; X64-NEXT:    paddq %xmm2, %xmm2
+; X64-NEXT:    psllq $31, %xmm2
+; X64-NEXT:    movq %xmm2, %rax
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm7
 ; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; X64-NEXT:    movq %xmm2, %rcx
-; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[2,3,2,3]
 ; X64-NEXT:    movq %xmm2, %rax
+; X64-NEXT:    movdqa %xmm1, %xmm2
+; X64-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-NEXT:    movq %xmm2, %rcx
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm2
 ; X64-NEXT:    punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm2[0]
 ; X64-NEXT:    movdqa {{.*#+}} xmm4 = [9223372039002259456,9223372039002259456]
 ; X64-NEXT:    movdqa %xmm7, %xmm2
 ; X64-NEXT:    pxor %xmm4, %xmm2
-; X64-NEXT:    movdqa {{.*#+}} xmm9 = [9223372043297226751,9223372043297226751]
-; X64-NEXT:    movdqa %xmm9, %xmm6
+; X64-NEXT:    movdqa {{.*#+}} xmm8 = [9223372043297226751,9223372043297226751]
+; X64-NEXT:    movdqa %xmm8, %xmm6
 ; X64-NEXT:    pcmpgtd %xmm2, %xmm6
-; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,0,2,2]
-; X64-NEXT:    pcmpeqd %xmm9, %xmm2
+; X64-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[0,0,2,2]
+; X64-NEXT:    pcmpeqd %xmm8, %xmm2
 ; X64-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[1,1,3,3]
-; X64-NEXT:    pand %xmm3, %xmm5
+; X64-NEXT:    pand %xmm9, %xmm5
 ; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[1,1,3,3]
 ; X64-NEXT:    por %xmm5, %xmm2
 ; X64-NEXT:    movdqa {{.*#+}} xmm6 = [8589934591,8589934591]
 ; X64-NEXT:    pand %xmm2, %xmm7
 ; X64-NEXT:    pandn %xmm6, %xmm2
 ; X64-NEXT:    por %xmm7, %xmm2
 ; X64-NEXT:    psrlq $1, %xmm2
-; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
-; X64-NEXT:    movq %xmm1, %rcx
-; X64-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; X64-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; X64-NEXT:    movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
+; X64-NEXT:    movq %xmm3, %rcx
 ; X64-NEXT:    paddq %xmm0, %xmm0
 ; X64-NEXT:    psllq $31, %xmm0
 ; X64-NEXT:    movq %xmm0, %rax
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm3
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X64-NEXT:    movq %xmm1, %rcx
 ; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; X64-NEXT:    movq %xmm0, %rax
+; X64-NEXT:    psrlq $32, %xmm1
+; X64-NEXT:    movq %xmm1, %rcx
 ; X64-NEXT:    xorl %edx, %edx
 ; X64-NEXT:    divq %rcx
 ; X64-NEXT:    movq %rax, %xmm0
 ; X64-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
 ; X64-NEXT:    pxor %xmm3, %xmm4
-; X64-NEXT:    movdqa %xmm9, %xmm0
+; X64-NEXT:    movdqa %xmm8, %xmm0
 ; X64-NEXT:    pcmpgtd %xmm4, %xmm0
 ; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,2,2]
-; X64-NEXT:    pcmpeqd %xmm9, %xmm4
+; X64-NEXT:    pcmpeqd %xmm8, %xmm4
 ; X64-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
 ; X64-NEXT:    pand %xmm1, %xmm4
 ; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; X64-NEXT:    por %xmm4, %xmm0
 ; X64-NEXT:    pand %xmm0, %xmm3
 ; X64-NEXT:    pandn %xmm6, %xmm0
 ; X64-NEXT:    por %xmm3, %xmm0
 ; X64-NEXT:    psrlq $1, %xmm0

llvm/test/CodeGen/X86/urem-seteq-vec-nonsplat.ll

 
 ; One all-ones divisor in even divisor
 define <4 x i32> @test_urem_even_allones_eq(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_even_allones_eq:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    psrld $31, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $31, %xmm2
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_allones_eq:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1

   ret <4 x i32> %ret
 }
 define <4 x i32> @test_urem_even_allones_ne(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_even_allones_ne:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    psrld $31, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $31, %xmm2
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pandn {{.*}}(%rip), %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_allones_ne:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1

 ; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_poweroftwo:

 
 ; One power-of-two divisor in even divisor
 define <4 x i32> @test_urem_even_poweroftwo(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_even_poweroftwo:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_poweroftwo:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1

 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,268435456,1374389535]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
 ; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
 ; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm3
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm2[0,1,2,3],xmm3[4,5,6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm3
-; CHECK-SSE41-NEXT:    psubd %xmm3, %xmm0
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
+; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_odd_even_poweroftwo:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
 ; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,268435456,1374389535]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm2, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm3, %xmm3
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;

 
 ; One one divisor in even divisor
 define <4 x i32> @test_urem_even_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_even_one:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psrld $2, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_one:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1

 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm3
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm3
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    movaps %xmm2, %xmm4
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [5,14,1,100]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; CHECK-SSE2-NEXT:    psrld $5, %xmm1
+; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm3
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; CHECK-SSE2-NEXT:    psubd %xmm4, %xmm0
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [5,14,1,100]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_even_one:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1

 ; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_INT_MIN:

 
 ; One INT_MIN divisor in even divisor
 define <4 x i32> @test_urem_even_INT_MIN(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_even_INT_MIN:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm3
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_INT_MIN:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1

 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE41-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [3435973837,2454267027,2,1374389535]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
 ; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm2
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
 ; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm3
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm2[0,1,2,3],xmm3[4,5,6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm3
-; CHECK-SSE41-NEXT:    psubd %xmm3, %xmm0
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
+; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_odd_even_INT_MIN:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
 ; CHECK-AVX1-NEXT:    vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [3435973837,2454267027,2,1374389535]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm2
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm2, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm3, %xmm3
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;

 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [5,4294967295,16,5]
+; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm3
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm2[2,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,4294967295,16,5]
-; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_allones_and_poweroftwo:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,3435973837]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm3
 ; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
 ; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm3
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm2[4,5,6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm3
-; CHECK-SSE41-NEXT:    psubd %xmm3, %xmm0
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
+; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_odd_allones_and_poweroftwo:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,3435973837]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm3, %xmm3
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;

 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [2454267027,2147483649,268435456,2454267027]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[1,2]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2,3,1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm2
+; CHECK-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [14,4294967295,16,14]
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm4
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm2[3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [14,4294967295,16,14]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm2
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_allones_and_poweroftwo:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3,4,5],xmm1[6,7]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [2454267027,2147483649,268435456,2454267027]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm4
 ; CHECK-SSE41-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3],xmm2[4,5],xmm4[6,7]
 ; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm4
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm2[4,5,6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm4 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm4
-; CHECK-SSE41-NEXT:    psubd %xmm4, %xmm0
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3],xmm2[4,5],xmm4[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
+; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_even_allones_and_poweroftwo:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3,4,5],xmm1[6,7]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [2454267027,2147483649,268435456,2454267027]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm4, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm3, %xmm3
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;

   ret <4 x i32> %ret
 }
 
 ; One all-ones divisor and power-of-two divisor divisor in odd+even divisor
 define <4 x i32> @test_urem_odd_even_allones_and_poweroftwo(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_even_allones_and_poweroftwo:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,1374389535]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm3
-; CHECK-SSE2-NEXT:    movdqa %xmm3, %xmm2
-; CHECK-SSE2-NEXT:    psrld $5, %xmm2
-; CHECK-SSE2-NEXT:    psrld $31, %xmm3
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,4294967295,16,100]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm4[2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,4294967295,16,100]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm3, %xmm4
+; CHECK-SSE2-NEXT:    psrld $5, %xmm4
+; CHECK-SSE2-NEXT:    psrld $31, %xmm3
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm4[3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_even_allones_and_poweroftwo:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,1374389535]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm3
 ; CHECK-SSE41-NEXT:    movdqa %xmm3, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $5, %xmm2
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm3
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm2[4,5,6,7]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm0, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm2
-; CHECK-SSE41-NEXT:    psrld $2, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE41-NEXT:    psrld $2, %xmm1
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm1
+; CHECK-SSE41-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_odd_even_allones_and_poweroftwo:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,1374389535]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $5, %xmm2, %xmm3
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[3,3,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;

 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [2454267027,2147483649,0,2454267027]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[1,2]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2,3,1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm4
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,0],xmm1[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [14,4294967295,1,14]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_allones_and_one:
 ; CHECK-SSE41:       # %bb.0:

 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,4294967295,1,100]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm4
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,2]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq

 ; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [5,16,1,5]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_poweroftwo_and_one:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,268435456,0,3435973837]

 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [2454267027,268435456,0,2454267027]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,3,2,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[1,2]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2,3,1]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    movaps %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[3,3]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [14,16,1,14]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_poweroftwo_and_one:
 ; CHECK-SSE41:       # %bb.0:

 ; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[3,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,16,1,100]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm4
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm0, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm4
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[3,0]
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,2]
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq

 ;------------------------------------------------------------------------------;
 
 define <4 x i32> @test_urem_odd_allones_and_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_odd_allones_and_poweroftwo_and_one:
 ; CHECK-SSE2:       # %bb.0:
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,0]
 ; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm3[2,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [5,4294967295,16,1]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm4
-; CHECK-SSE2-NEXT:    psrld $31, %xmm4
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE2-NEXT:    movdqa %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    psrld $2, %xmm1
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [5,4294967295,16,1]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; CHECK-SSE2-NEXT:    psrld $31, %xmm3
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    psubd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_odd_allones_and_poweroftwo_and_one:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,0]
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm2
 ; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm2
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[3,3,3,3]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    movdqa %xmm2, %xmm3
-; CHECK-SSE41-NEXT:    psrld $2, %xmm3
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; CHECK-SSE41-NEXT:    psrld $2, %xmm2
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE41-NEXT:    psrld $31, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3],xmm3[4,5],xmm2[6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm0[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm3
+; CHECK-SSE41-NEXT:    psrld $31, %xmm3
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm0[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm3
+; CHECK-SSE41-NEXT:    psubd %xmm3, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_odd_allones_and_poweroftwo_and_one:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3435973837,2147483649,268435456,0]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[3,3,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm2, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm2, %xmm2
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm3[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm1, %xmm3, %xmm1
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm0[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0

   %cmp = icmp eq <4 x i32> %urem, <i32 0, i32 0, i32 0, i32 0>
   %ret = zext <4 x i1> %cmp to <4 x i32>
   ret <4 x i32> %ret
 }
 
 define <4 x i32> @test_urem_even_allones_and_poweroftwo_and_one(<4 x i32> %X) nounwind {
 ; CHECK-SSE2-LABEL: test_urem_even_allones_and_poweroftwo_and_one:
 ; CHECK-SSE2:       # %bb.0:
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm1
-; CHECK-SSE2-NEXT:    psrld $1, %xmm1
-; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
 ; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2454267027,2147483649,268435456,0]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,3,2,3]
+; CHECK-SSE2-NEXT:    movdqa %xmm0, %xmm4
+; CHECK-SSE2-NEXT:    psrld $1, %xmm4
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,3]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm4
+; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,3,2,3]
+; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; CHECK-SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; CHECK-SSE2-NEXT:    psrld $2, %xmm2
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm3[2,3]
-; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [14,4294967295,16,1]
-; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
+; CHECK-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [14,4294967295,16,1]
+; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm2
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT:    psrld $31, %xmm3
+; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm0[3,3]
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm1, %xmm4
-; CHECK-SSE2-NEXT:    psrld $31, %xmm4
-; CHECK-SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm0[3,3]
-; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
-; CHECK-SSE2-NEXT:    pmuludq %xmm4, %xmm1
+; CHECK-SSE2-NEXT:    pmuludq %xmm3, %xmm1
 ; CHECK-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; CHECK-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; CHECK-SSE2-NEXT:    psubd %xmm2, %xmm0
 ; CHECK-SSE2-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE2-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE2-NEXT:    retq
 ;
 ; CHECK-SSE41-LABEL: test_urem_even_allones_and_poweroftwo_and_one:
 ; CHECK-SSE41:       # %bb.0:
 ; CHECK-SSE41-NEXT:    movdqa %xmm0, %xmm1
 ; CHECK-SSE41-NEXT:    psrld $1, %xmm1
 ; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3,4,5,6,7]
 ; CHECK-SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [2454267027,2147483649,268435456,0]
 ; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm1
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,3,3,3]
 ; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT:    movdqa %xmm1, %xmm3
-; CHECK-SSE41-NEXT:    psrld $2, %xmm3
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm1[4,5,6,7]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT:    pmuludq %xmm1, %xmm2
-; CHECK-SSE41-NEXT:    psrld $31, %xmm2
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3],xmm3[4,5],xmm2[6,7]
-; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm0[6,7]
-; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT:    psubd %xmm2, %xmm0
+; CHECK-SSE41-NEXT:    psrld $2, %xmm1
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; CHECK-SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; CHECK-SSE41-NEXT:    pmuludq %xmm2, %xmm3
+; CHECK-SSE41-NEXT:    psrld $31, %xmm3
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
+; CHECK-SSE41-NEXT:    pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm0[6,7]
+; CHECK-SSE41-NEXT:    pmulld {{.*}}(%rip), %xmm3
+; CHECK-SSE41-NEXT:    psubd %xmm3, %xmm0
 ; CHECK-SSE41-NEXT:    pxor %xmm1, %xmm1
 ; CHECK-SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
 ; CHECK-SSE41-NEXT:    psrld $31, %xmm0
 ; CHECK-SSE41-NEXT:    retq
 ;
 ; CHECK-AVX1-LABEL: test_urem_even_allones_and_poweroftwo_and_one:
 ; CHECK-AVX1:       # %bb.0:
 ; CHECK-AVX1-NEXT:    vpsrld $1, %xmm0, %xmm1
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3,4,5,6,7]
 ; CHECK-AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [2454267027,2147483649,268435456,0]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
+; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[3,3,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm3
-; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm3[0,1,2,3],xmm1[4,5,6,7]
+; CHECK-AVX1-NEXT:    vpsrld $2, %xmm1, %xmm1
+; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
 ; CHECK-AVX1-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
 ; CHECK-AVX1-NEXT:    vpsrld $31, %xmm2, %xmm2
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm0[6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0

llvm/test/CodeGen/X86/urem-seteq-vec-nonzero.ll

 ; CHECK-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7]
 ; CHECK-AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    vpcmpeqd {{.*}}(%rip), %xmm0, %xmm0
 ; CHECK-AVX1-NEXT:    retq
 ;
 ; CHECK-AVX2-LABEL: t32_tautological:
 ; CHECK-AVX2:       # %bb.0:
-; CHECK-AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,0,2147483648,2863311531]
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm3, %xmm2
-; CHECK-AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm1
-; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; CHECK-AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; CHECK-AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [2863311531,2863311531,2863311531,2863311531]
+; CHECK-AVX2-NEXT:    vpmuludq %xmm2, %xmm1, %xmm1
+; CHECK-AVX2-NEXT:    vpmuludq {{.*}}(%rip), %xmm0, %xmm2
+; CHECK-AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; CHECK-AVX2-NEXT:    vpsrlvd {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3]
 ; CHECK-AVX2-NEXT:    vpmulld {{.*}}(%rip), %xmm1, %xmm1
 ; CHECK-AVX2-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
 ; CHECK-AVX2-NEXT:    vpcmpeqd {{.*}}(%rip), %xmm0, %xmm0
 ; CHECK-AVX2-NEXT:    retq
 ;
 ; CHECK-AVX512VL-LABEL: t32_tautological:

llvm/test/CodeGen/X86/vec_insert-2.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse2,-sse4.1 | FileCheck %s --check-prefix=X32
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,-sse4.1 | FileCheck %s --check-prefix=X64
 
 define <4 x float> @t1(float %s, <4 x float> %tmp) nounwind {
 ; X32-LABEL: t1:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; X32-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: t1:
 ; X64:       # %bb.0:
-; X64-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[2,0]
+; X64-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; X64-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; X64-NEXT:    movaps %xmm1, %xmm0
 ; X64-NEXT:    retq
   %tmp1 = insertelement <4 x float> %tmp, float %s, i32 3
   ret <4 x float> %tmp1
 }
 
 define <4 x i32> @t2(i32 %s, <4 x i32> %tmp) nounwind {
 ; X32-LABEL: t2:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; X32-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: t2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movd %edi, %xmm1
-; X64-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; X64-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; X64-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; X64-NEXT:    retq
   %tmp1 = insertelement <4 x i32> %tmp, i32 %s, i32 3
   ret <4 x i32> %tmp1
 }
 
 define <2 x double> @t3(double %s, <2 x double> %tmp) nounwind {
 ; X32-LABEL: t3:

llvm/test/CodeGen/X86/vec_insert-3.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse2,-sse4.1 | FileCheck %s --check-prefix=X32
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,-sse4.1 | FileCheck %s --check-prefix=X64
 
 define <2 x i64> @t1(i64 %s, <2 x i64> %tmp) nounwind {
 ; X32-LABEL: t1:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X32-NEXT:    movaps %xmm0, %xmm2
-; X32-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; X32-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm2[2,0]
-; X32-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
+; X32-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; X32-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; X32-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: t1:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq %rdi, %xmm1
 ; X64-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; X64-NEXT:    retq
   %tmp1 = insertelement <2 x i64> %tmp, i64 %s, i32 1
   ret <2 x i64> %tmp1
 }

llvm/test/CodeGen/X86/vec_insert-5.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i386-unknown -mattr=+sse2,+ssse3 | FileCheck %s --check-prefix=X32
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,+ssse3 | FileCheck %s --check-prefix=X64
 
 ; There are no MMX operations in @t1
 
 define void  @t1(i32 %a, x86_mmx* %P) nounwind {
 ; X32-LABEL: t1:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X32-NEXT:    shll $12, %ecx
 ; X32-NEXT:    movd %ecx, %xmm0
-; X32-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; X32-NEXT:    psllq $32, %xmm0
 ; X32-NEXT:    movq %xmm0, (%eax)
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: t1:
 ; X64:       # %bb.0:
 ; X64-NEXT:    shll $12, %edi
 ; X64-NEXT:    movd %edi, %xmm0
-; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,1,1]
+; X64-NEXT:    psllq $32, %xmm0
 ; X64-NEXT:    movq %xmm0, (%rsi)
 ; X64-NEXT:    retq
  %tmp12 = shl i32 %a, 12
  %tmp21 = insertelement <2 x i32> undef, i32 %tmp12, i32 1
  %tmp22 = insertelement <2 x i32> %tmp21, i32 0, i32 0
  %tmp23 = bitcast <2 x i32> %tmp22 to x86_mmx
  store x86_mmx %tmp23, x86_mmx* %P
  ret void
 }
 
 define <4 x float> @t2(<4 x float>* %P) nounwind {
 ; X32-LABEL: t2:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    xorps %xmm0, %xmm0
 ; X32-NEXT:    xorps %xmm1, %xmm1
-; X32-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],mem[0,0]
+; X32-NEXT:    unpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; X32-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: t2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorps %xmm0, %xmm0
 ; X64-NEXT:    xorps %xmm1, %xmm1
-; X64-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],mem[0,0]
+; X64-NEXT:    unpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; X64-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; X64-NEXT:    retq
   %tmp1 = load <4 x float>, <4 x float>* %P
   %tmp2 = shufflevector <4 x float> %tmp1, <4 x float> zeroinitializer, <4 x i32> < i32 4, i32 4, i32 4, i32 0 >
   ret <4 x float> %tmp2
 }
 
 define <4 x float> @t3(<4 x float>* %P) nounwind {

llvm/test/CodeGen/X86/vec_int_to_fp.ll

   %cvt = sitofp <4 x i32> %a to <4 x double>
   %shuf = shufflevector <4 x double> %cvt, <4 x double> undef, <2 x i32> <i32 0, i32 1>
   ret <2 x double> %shuf
 }
 
 define <2 x double> @sitofp_2i16_to_2f64(<8 x i16> %a) {
 ; SSE2-LABEL: sitofp_2i16_to_2f64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    psrad $16, %xmm0
 ; SSE2-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: sitofp_2i16_to_2f64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0
 ; SSE41-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: sitofp_2i16_to_2f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpmovsxwd %xmm0, %xmm0
 ; AVX-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %shuf = shufflevector <8 x i16> %a, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
   %cvt = sitofp <2 x i16> %shuf to <2 x double>
   ret <2 x double> %cvt
 }
 
 define <2 x double> @sitofp_8i16_to_2f64(<8 x i16> %a) {
 ; SSE2-LABEL: sitofp_8i16_to_2f64:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    psrad $16, %xmm0
 ; SSE2-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: sitofp_8i16_to_2f64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovsxwd %xmm0, %xmm0
 ; SSE41-NEXT:    cvtdq2pd %xmm0, %xmm0

   %shuf = shufflevector <8 x double> %cvt, <8 x double> undef, <2 x i32> <i32 0, i32 1>
   ret <2 x double> %shuf
 }
 
 define <2 x double> @sitofp_2i8_to_2f64(<16 x i8> %a) {
 ; SSE2-LABEL: sitofp_2i8_to_2f64:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    psrad $24, %xmm0
 ; SSE2-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: sitofp_2i8_to_2f64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovsxbd %xmm0, %xmm0
 ; SSE41-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: sitofp_2i8_to_2f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpmovsxbd %xmm0, %xmm0
 ; AVX-NEXT:    vcvtdq2pd %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %shuf = shufflevector <16 x i8> %a, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
   %cvt = sitofp <2 x i8> %shuf to <2 x double>
   ret <2 x double> %cvt
 }
 
 define <2 x double> @sitofp_16i8_to_2f64(<16 x i8> %a) {
 ; SSE2-LABEL: sitofp_16i8_to_2f64:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    psrad $24, %xmm0
 ; SSE2-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: sitofp_16i8_to_2f64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovsxbd %xmm0, %xmm0
 ; SSE41-NEXT:    cvtdq2pd %xmm0, %xmm0

 }
 
 define <2 x double> @sitofp_load_2i8_to_2f64(<2 x i8> *%a) {
 ; SSE2-LABEL: sitofp_load_2i8_to_2f64:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movzwl (%rdi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE2-NEXT:    psrad $24, %xmm0
 ; SSE2-NEXT:    cvtdq2pd %xmm0, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: sitofp_load_2i8_to_2f64:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movzwl (%rdi), %eax
 ; SSE41-NEXT:    movd %eax, %xmm0

llvm/test/CodeGen/X86/vec_set-6.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i386-unknown -mattr=+sse2,+sse4.1 | FileCheck %s --check-prefix=X86
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2,+sse4.1 | FileCheck %s --check-prefix=X64
 
 define <4 x float> @test(float %a, float %b, float %c) nounwind {
 ; X86-LABEL: test:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movsd {{.*#+}} xmm1 = mem[0],zero
 ; X86-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; X86-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,1]
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: test:
 ; X64:       # %bb.0:
-; X64-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-NEXT:    insertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],zero,zero
 ; X64-NEXT:    xorps %xmm2, %xmm2
 ; X64-NEXT:    blendps {{.*#+}} xmm0 = xmm0[0],xmm2[1,2,3]
 ; X64-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,1]
 ; X64-NEXT:    retq
   %tmp = insertelement <4 x float> zeroinitializer, float %a, i32 1
   %tmp8 = insertelement <4 x float> %tmp, float %b, i32 2
   %tmp10 = insertelement <4 x float> %tmp8, float %c, i32 3
   ret <4 x float> %tmp10
 }
 

llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll

 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm3, %xmm3
 ; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm3
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm2, %xmm2
 ; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
 ; AVX1-NEXT:    vpackuswb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm4
 ; AVX1-NEXT:    vpsllw $7, %xmm4, %xmm4
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
 ; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm3
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm3, %xmm3
 ; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm3
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm4 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm5 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; AVX1-NEXT:    vpsllw $7, %xmm5, %xmm5

 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm4, %xmm4
 ; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm4, %xmm4
 ; AVX1-NEXT:    vpand %xmm2, %xmm4, %xmm4
 ; AVX1-NEXT:    vpackuswb %xmm5, %xmm4, %xmm4
 ; AVX1-NEXT:    vpsubb %xmm4, %xmm3, %xmm3
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm5
 ; AVX1-NEXT:    vpsllw $7, %xmm5, %xmm5
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6],xmm5[7]
 ; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm4, %xmm4
 ; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
 ; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm5 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; AVX1-NEXT:    vpsllw $7, %xmm6, %xmm6

llvm/test/CodeGen/X86/vector-pack-256.ll

 define <16 x i16> @concat_trunc_packusdw_256(<8 x i32> %a0, <8 x i32> %a1) nounwind {
 ; AVX1-LABEL: concat_trunc_packusdw_256:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; AVX1-NEXT:    vpsrld $17, %xmm2, %xmm2
 ; AVX1-NEXT:    vpsrld $17, %xmm0, %xmm0
 ; AVX1-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm2 = xmm0[1],xmm1[1]
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    retq

llvm/test/CodeGen/X86/vector-reduce-and-bool.ll

 ; SSE41-NEXT:    pmovmskb %xmm0, %eax
 ; SSE41-NEXT:    cmpb $-1, %al
 ; SSE41-NEXT:    sete %al
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_v8i32_v8i1:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpsllw $15, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    cmpb $-1, %al
 ; AVX1-NEXT:    sete %al

llvm/test/CodeGen/X86/vector-reduce-mul.ll

 
 define i64 @test_v2i64(<2 x i64> %a0) {
 ; SSE-LABEL: test_v2i64:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    psrlq $32, %xmm2
 ; SSE-NEXT:    pmuludq %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; SSE-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    movq %xmm0, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: test_v2i64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vmovq %xmm0, %rax
 ; AVX-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_v2i64:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BW-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BW-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vmovq %xmm0, %rax
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512BWVL-LABEL: test_v2i64:
 ; AVX512BWVL:       # %bb.0:
 ; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BWVL-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BWVL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vmovq %xmm0, %rax
 ; AVX512BWVL-NEXT:    retq
 ;

 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    psrlq $32, %xmm2
 ; SSE-NEXT:    pmuludq %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; SSE-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    movq %xmm0, %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: test_v4i64:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
 ; AVX1-NEXT:    vpsrlq $32, %xmm1, %xmm3
 ; AVX1-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: test_v4i64:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
 ; AVX2-NEXT:    vpsrlq $32, %xmm1, %xmm3
 ; AVX2-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX2-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_v4i64:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX512BW-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpsrlq $32, %xmm1, %xmm3
 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BW-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BW-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vmovq %xmm0, %rax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512BWVL-LABEL: test_v4i64:
 ; AVX512BWVL:       # %bb.0:
 ; AVX512BWVL-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX512BWVL-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpsrlq $32, %xmm1, %xmm3
 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BWVL-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BWVL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vmovq %xmm0, %rax
 ; AVX512BWVL-NEXT:    vzeroupper
 ; AVX512BWVL-NEXT:    retq

 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    psrlq $32, %xmm2
 ; SSE-NEXT:    pmuludq %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; SSE-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    movq %xmm0, %rax
 ; SSE-NEXT:    retq
 ;

 ; AVX1-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm3, %xmm1
 ; AVX1-NEXT:    vpsllq $32, %xmm1, %xmm1
 ; AVX1-NEXT:    vpmuludq %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq

 ; AVX2-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX2-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq

 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BW-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BW-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vmovq %xmm0, %rax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq

 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BWVL-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BWVL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vmovq %xmm0, %rax
 ; AVX512BWVL-NEXT:    vzeroupper
 ; AVX512BWVL-NEXT:    retq

 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    psrlq $32, %xmm2
 ; SSE-NEXT:    pmuludq %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; SSE-NEXT:    pmuludq %xmm0, %xmm3
 ; SSE-NEXT:    paddq %xmm2, %xmm3
 ; SSE-NEXT:    psllq $32, %xmm3
 ; SSE-NEXT:    pmuludq %xmm1, %xmm0
 ; SSE-NEXT:    paddq %xmm3, %xmm0
 ; SSE-NEXT:    movq %xmm0, %rax
 ; SSE-NEXT:    retq
 ;

 ; AVX1-NEXT:    vpmuludq %xmm3, %xmm1, %xmm3
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm0, %xmm1, %xmm0
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX1-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovq %xmm0, %rax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq

 ; AVX2-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX2-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX2-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovq %xmm0, %rax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq

 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BW-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BW-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BW-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BW-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BW-NEXT:    vmovq %xmm0, %rax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq

 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX512BWVL-NEXT:    vpsrlq $32, %xmm0, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm2, %xmm2
-; AVX512BWVL-NEXT:    vpsrldq {{.*#+}} xmm3 = xmm0[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BWVL-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
 ; AVX512BWVL-NEXT:    vpmuludq %xmm3, %xmm0, %xmm3
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm3, %xmm2
 ; AVX512BWVL-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512BWVL-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vpaddq %xmm2, %xmm0, %xmm0
 ; AVX512BWVL-NEXT:    vmovq %xmm0, %rax
 ; AVX512BWVL-NEXT:    vzeroupper
 ; AVX512BWVL-NEXT:    retq

 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    pand %xmm1, %xmm2
 ; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    packuswb %xmm3, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pmullw %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    packuswb %xmm3, %xmm0
-; SSE2-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm2, %xmm0
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    packuswb %xmm1, %xmm1
-; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    pmullw %xmm0, %xmm1
-; SSE2-NEXT:    movd %xmm1, %eax
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    packuswb %xmm3, %xmm1
+; SSE2-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pmullw %xmm2, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psrld $8, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v16i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE41-NEXT:    pmullw %xmm1, %xmm0

 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    pand %xmm1, %xmm2
 ; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    packuswb %xmm3, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pmullw %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    packuswb %xmm3, %xmm0
-; SSE2-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm2, %xmm0
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    packuswb %xmm1, %xmm1
-; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    pmullw %xmm0, %xmm1
-; SSE2-NEXT:    movd %xmm1, %eax
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    packuswb %xmm3, %xmm1
+; SSE2-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pmullw %xmm2, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psrld $8, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v32i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
 ; SSE41-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero

 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    pand %xmm1, %xmm2
 ; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    packuswb %xmm3, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pmullw %xmm0, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    packuswb %xmm3, %xmm0
-; SSE2-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm2, %xmm0
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    packuswb %xmm1, %xmm1
-; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    pmullw %xmm0, %xmm1
-; SSE2-NEXT:    movd %xmm1, %eax
+; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    packuswb %xmm3, %xmm1
+; SSE2-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pmullw %xmm2, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psrld $8, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v64i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm4 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
 ; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm5 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero

 ; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
 ; SSE2-NEXT:    pand %xmm0, %xmm2
 ; SSE2-NEXT:    pxor %xmm3, %xmm3
 ; SSE2-NEXT:    packuswb %xmm3, %xmm2
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,2,3,3]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pmullw %xmm1, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, %xmm1
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    packuswb %xmm3, %xmm1
-; SSE2-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm2, %xmm1
-; SSE2-NEXT:    pand %xmm1, %xmm0
-; SSE2-NEXT:    packuswb %xmm0, %xmm0
-; SSE2-NEXT:    psrlw $8, %xmm0
-; SSE2-NEXT:    pmullw %xmm1, %xmm0
-; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    packuswb %xmm3, %xmm0
+; SSE2-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pmullw %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrld $8, %xmm1
+; SSE2-NEXT:    psrlw $8, %xmm1
+; SSE2-NEXT:    pmullw %xmm0, %xmm1
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_v128i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm8 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero
 ; SSE41-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm9 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero

llvm/test/CodeGen/X86/vector-reduce-or-bool.ll

 ; SSE41-NEXT:    pmovmskb %xmm0, %eax
 ; SSE41-NEXT:    testl $43690, %eax # imm = 0xAAAA
 ; SSE41-NEXT:    setne %al
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_v8i32_v8i1:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpsllw $15, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    testl $43690, %eax # imm = 0xAAAA
 ; AVX1-NEXT:    setne %al
 ; AVX1-NEXT:    vzeroupper

llvm/test/CodeGen/X86/vector-reduce-xor-bool.ll

 ; SSE41-NEXT:    pmovmskb %xmm0, %eax
 ; SSE41-NEXT:    testb %al, %al
 ; SSE41-NEXT:    setnp %al
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_v8i32_v8i1:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpsllw $15, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
 ; AVX1-NEXT:    testb %al, %al
 ; AVX1-NEXT:    setnp %al

llvm/test/CodeGen/X86/vector-shuffle-128-v16.ll

 ; XOP-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> <i8 0, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef>, <16 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06(<16 x i8> %a) {
 ; SSE-LABEL: shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero
+; SSE-NEXT:    psrlq $8, %xmm0
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero
+; AVX1-NEXT:    vpsrlq $8, %xmm0, %xmm0
 ; AVX1-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero
+; AVX2-SLOW-NEXT:    vpsrlq $8, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[1,2,3,4,5,6]
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vpshufb {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[1,2,3,4,5,6]
 ; AVX512VL-NEXT:    retq
 ;
 ; XOP-LABEL: shuffle_v16i8_zz_zz_zz_zz_zz_zz_zz_zz_zz_zz_01_02_03_04_05_06:
 ; XOP:       # %bb.0:
-; XOP-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero
+; XOP-NEXT:    vpsrlq $8, %xmm0, %xmm0
 ; XOP-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; XOP-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> <i8 0, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef, i8 undef>, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6>
   ret <16 x i8> %shuffle
 }
 
 define <16 x i8> @shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14(<16 x i8> %a) {
 ; SSE2-LABEL: shuffle_v16i8_03_00_01_02_07_04_05_06_11_08_09_10_15_12_13_14:

 
 define <16 x i8> @PR31301(i8* nocapture readonly %x, i8* nocapture readonly %y)  {
 ; SSE2-LABEL: PR31301:
 ; SSE2:       # %bb.0: # %entry
 ; SSE2-NEXT:    movzbl (%rdi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm0
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE2-NEXT:    movzbl (%rsi), %eax
 ; SSE2-NEXT:    movd %eax, %xmm1
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: PR31301:
 ; SSSE3:       # %bb.0: # %entry
 ; SSSE3-NEXT:    movzbl (%rdi), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm0
 ; SSSE3-NEXT:    pxor %xmm1, %xmm1

llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll

 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 6, i32 7, i32 2, i32 3>
   ret <4 x float> %shuffle
 }
 
 define <4 x i32> @shuffle_v4i32_0124(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-LABEL: shuffle_v4i32_0124:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_0124:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_0124:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4i32_0124:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
 ; SSE41-NEXT:    retq

 ; AVX512VL-NEXT:    vpermt2d %xmm1, %xmm2, %xmm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 1, i32 2, i32 4>
   ret <4 x i32> %shuffle
 }
 define <4 x i32> @shuffle_v4i32_0142(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-LABEL: shuffle_v4i32_0142:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_0142:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_0142:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4i32_0142:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5],xmm0[6,7]

 ; AVX512VL-NEXT:    vpermt2d %xmm1, %xmm2, %xmm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 1, i32 4, i32 2>
   ret <4 x i32> %shuffle
 }
 define <4 x i32> @shuffle_v4i32_0412(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-LABEL: shuffle_v4i32_0412:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[1,2]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_0412:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[1,2]
 ; SSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_0412:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[1,2]
 ; SSSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4i32_0412:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,2]
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
 ; SSE41-NEXT:    retq
 ;
-; AVX1OR2-LABEL: shuffle_v4i32_0412:
-; AVX1OR2:       # %bb.0:
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,2]
-; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
-; AVX1OR2-NEXT:    retq
+; AVX1-LABEL: shuffle_v4i32_0412:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,2]
+; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shuffle_v4i32_0412:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,1,2]
+; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i32_0412:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,4,1,2]
 ; AVX512VL-NEXT:    vpermt2d %xmm1, %xmm2, %xmm0
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 4, i32 1, i32 2>
   ret <4 x i32> %shuffle
 }
 define <4 x i32> @shuffle_v4i32_4012(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-LABEL: shuffle_v4i32_4012:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[1,2]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_4012:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[1,2]
 ; SSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_4012:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[0,0]
+; SSSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[1,2]
 ; SSSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4i32_4012:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,2]
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3,4,5,6,7]

 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 4, i32 1, i32 2, i32 3>
   ret <4 x float> %shuffle
 }
 
 define <4 x float> @shuffle_v4f32_z4zz(<4 x float> %a) {
 ; SSE2-LABEL: shuffle_v4f32_z4zz:
 ; SSE2:       # %bb.0:
+; SSE2-NEXT:    movq {{.*#+}} xmm0 = xmm0[0],zero
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4f32_z4zz:
 ; SSE3:       # %bb.0:
+; SSE3-NEXT:    movq {{.*#+}} xmm0 = xmm0[0],zero
 ; SSE3-NEXT:    xorps %xmm1, %xmm1
-; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4f32_z4zz:
 ; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movq {{.*#+}} xmm0 = xmm0[0],zero
 ; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4f32_z4zz:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = zero,xmm0[0],zero,zero
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v4f32_z4zz:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm0[0],zero,zero
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 2, i32 4, i32 3, i32 0>
   ret <4 x float> %shuffle
 }
 
 define <4 x float> @shuffle_v4f32_zz4z(<4 x float> %a) {
 ; SSE2-LABEL: shuffle_v4f32_zz4z:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[3,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE2-NEXT:    movaps %xmm1, %xmm0
+; SSE2-NEXT:    movq {{.*#+}} xmm1 = xmm0[0],zero
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4f32_zz4z:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    xorps %xmm1, %xmm1
-; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[3,0]
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE3-NEXT:    movaps %xmm1, %xmm0
+; SSE3-NEXT:    movq {{.*#+}} xmm1 = xmm0[0],zero
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4f32_zz4z:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[3,0]
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSSE3-NEXT:    movaps %xmm1, %xmm0
+; SSSE3-NEXT:    movq {{.*#+}} xmm1 = xmm0[0],zero
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4f32_zz4z:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = zero,zero,xmm0[0],zero
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v4f32_zz4z:

   %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 0, i32 1, i32 2, i32 7>
   ret <4 x float> %shuffle
 }
 
 define <4 x float> @shuffle_v4f32_z6zz(<4 x float> %a) {
 ; SSE2-LABEL: shuffle_v4f32_z6zz:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
+; SSE2-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4f32_z6zz:
 ; SSE3:       # %bb.0:
 ; SSE3-NEXT:    xorps %xmm1, %xmm1
-; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
+; SSE3-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4f32_z6zz:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
+; SSSE3-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4f32_z6zz:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = zero,xmm0[2],zero,zero
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v4f32_z6zz:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm0[2],zero,zero
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> zeroinitializer, <4 x float> %a, <4 x i32> <i32 0, i32 6, i32 2, i32 3>
   ret <4 x float> %shuffle
 }
 
 define <4 x float> @shuffle_v4f32_0z23(<4 x float> %a) {
 ; SSE2-LABEL: shuffle_v4f32_0z23:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4f32_0z23:
 ; SSE3:       # %bb.0:
 ; SSE3-NEXT:    xorps %xmm1, %xmm1
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; SSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4f32_0z23:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,3]
 ; SSSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4f32_0z23:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    xorps %xmm1, %xmm1
 ; SSE41-NEXT:    blendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]

 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 undef, i32 0, i32 5, i32 1>
   ret <4 x float> %shuffle
 }
 
 define <4 x float> @shuffle_v4f32_0zz4(<4 x float> %a, <4 x float> %b) {
 ; SSE2-LABEL: shuffle_v4f32_0zz4:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorps %xmm2, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm2[2,0]
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE2-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE2-NEXT:    movaps %xmm2, %xmm0
+; SSE2-NEXT:    movq {{.*#+}} xmm2 = xmm1[0],zero
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
+; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4f32_0zz4:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    xorps %xmm2, %xmm2
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm2[2,0]
-; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE3-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE3-NEXT:    movaps %xmm2, %xmm0
+; SSE3-NEXT:    movq {{.*#+}} xmm2 = xmm1[0],zero
+; SSE3-NEXT:    pxor %xmm1, %xmm1
+; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
+; SSE3-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; SSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4f32_0zz4:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    xorps %xmm2, %xmm2
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm2[2,0]
-; SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSSE3-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSSE3-NEXT:    movaps %xmm2, %xmm0
+; SSSE3-NEXT:    movq {{.*#+}} xmm2 = xmm1[0],zero
+; SSSE3-NEXT:    pxor %xmm1, %xmm1
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
+; SSSE3-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; SSSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4f32_0zz4:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],zero,zero,xmm1[0]
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v4f32_0zz4:

   %shuffle = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32> <i32 0, i32 undef, i32 undef, i32 6>
   %shuffle1 = shufflevector <4 x float> zeroinitializer, <4 x float> %shuffle, <4 x i32> <i32 4, i32 1, i32 2, i32 7>
   ret <4 x float> %shuffle1
 }
 
 define <4 x float> @shuffle_v4f32_0z24(<4 x float> %a, <4 x float> %b) {
 ; SSE2-LABEL: shuffle_v4f32_0z24:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    xorps %xmm2, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE2-NEXT:    movaps %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4f32_0z24:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE3-NEXT:    xorps %xmm2, %xmm2
-; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSE3-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE3-NEXT:    movaps %xmm2, %xmm0
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4f32_0z24:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSSE3-NEXT:    xorps %xmm2, %xmm2
-; SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSSE3-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSSE3-NEXT:    movaps %xmm2, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4f32_0z24:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0],zero,xmm0[2],xmm1[0]
 ; SSE41-NEXT:    retq

 ; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> zeroinitializer, <4 x i32> %a, <4 x i32> <i32 0, i32 undef, i32 undef, i32 4>
   ret <4 x i32> %shuffle
 }
 
 define <4 x i32> @shuffle_v4i32_z6zz(<4 x i32> %a) {
 ; SSE2-LABEL: shuffle_v4i32_z6zz:
 ; SSE2:       # %bb.0:
+; SSE2-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_z6zz:
 ; SSE3:       # %bb.0:
+; SSE3-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE3-NEXT:    xorps %xmm1, %xmm1
-; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_z6zz:
 ; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
 ; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4i32_z6zz:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,2,3,3]
 ; SSE41-NEXT:    pxor %xmm0, %xmm0
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]

   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
   ret <4 x i32> %shuffle
 }
 
 ; PR22391
 define <4 x i32> @shuffle_v4i32_2456(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-LABEL: shuffle_v4i32_2456:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm1[0,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[0,1]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[1,2]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_2456:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm1[0,0]
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[0,1]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[1,2]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_2456:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
 ; SSSE3-NEXT:    palignr {{.*#+}} xmm1 = xmm0[12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10,11]
 ; SSSE3-NEXT:    movdqa %xmm1, %xmm0

   %1 = extractelement <4 x i32> %a1, i32 3
   %2 = insertelement <4 x i32> %a0, i32 %1, i32 0
   ret <4 x i32> %2
 }
 
 define <4 x i32> @extract3_insert3_v4i32_0127(<4 x i32> %a0, <4 x i32> %a1) {
 ; SSE2-LABEL: extract3_insert3_v4i32_0127:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: extract3_insert3_v4i32_0127:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[2,0]
+; SSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: extract3_insert3_v4i32_0127:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[2,0]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: extract3_insert3_v4i32_0127:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    blendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
 ; SSE41-NEXT:    retq
 ;

llvm/test/CodeGen/X86/vector-shuffle-128-v8.ll

   %shuffle = shufflevector <8 x i16> %a, <8 x i16> %b, <8 x i32> <i32 4, i32 4, i32 3, i32 10, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i16> %shuffle
 }
 
 define <8 x i16> @shuffle_v8i16_032dXXXX(<8 x i16> %a, <8 x i16> %b) {
 ; SSE2-LABEL: shuffle_v8i16_032dXXXX:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1,2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v8i16_032dXXXX:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,xmm1[10,11,u,u,u,u,u,u,u,u]

 
   ret <8 x i16> %shuffle
 }
 
 ; PR40306
 define <8 x i16> @shuffle_v8i16_9zzzuuuu(<8 x i16> %x) {
 ; SSE-LABEL: shuffle_v8i16_9zzzuuuu:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: shuffle_v8i16_9zzzuuuu:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i16_9zzzuuuu:
 ; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3]
+; AVX2-SLOW-NEXT:    vpbroadcastd %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: shuffle_v8i16_9zzzuuuu:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-SLOW-LABEL: shuffle_v8i16_9zzzuuuu:
 ; AVX512VL-SLOW:       # %bb.0:
-; AVX512VL-SLOW-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3]
+; AVX512VL-SLOW-NEXT:    vbroadcastss %xmm0, %xmm0
 ; AVX512VL-SLOW-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX512VL-SLOW-NEXT:    retq
 ;
 ; AVX512VL-FAST-LABEL: shuffle_v8i16_9zzzuuuu:
 ; AVX512VL-FAST:       # %bb.0:
 ; AVX512VL-FAST-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[2,3],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; AVX512VL-FAST-NEXT:    retq
 ;

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

 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,0,1,4,5,8,9,14,15]
 ; AVX1-NEXT:    vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-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:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_05_06_07_16_17_18_27_12_13_14_15_24_25_26_27:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [4,5,6,7,16,17,18,27,12,13,14,15,24,25,26,27]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_04_05_06_07_16_17_18_27_12_13_14_15_24_25_26_27:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; XOPAVX1-NEXT:    vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,4,5,8,9,14,15]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_04_05_06_07_16_17_18_27_12_13_14_15_24_25_26_27:
 ; XOPAVX2:       # %bb.0:
 ; XOPAVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; XOPAVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; XOPAVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; XOPAVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; XOPAVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[2,3,0,1,6,7,4,5]
 ; XOPAVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 27, i32 12, i32 13, i32 14, i32 15, i32 24, i32 25, i32 26, i32 27>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_00_00_00_00_00_00_08_08_08_08_08_08_08_08_08(<16 x i16> %a, <16 x i16> %b) {

 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,4,5,8,9,14,15,u,u,u,u,u,u,u,u]
 ; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    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:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
 ; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,2,3,10,11,4,5,12,13,6,7,u,u,16,17,24,25,18,19,26,27,20,21,28,29,22,23,u,u]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; 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:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,20,1,21,2,22,3,31,8,28,9,29,10,30,11,31]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_00_20_01_21_02_22_03_31_08_28_09_29_10_30_11_31:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; XOPAVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,3,2,3]
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13],xmm3[14,15],xmm1[u,u,u,u,u,u,u,u]
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_20_01_21_02_22_03_31_08_28_09_29_10_30_11_31:
 ; XOPAVX2:       # %bb.0:
 ; XOPAVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; XOPAVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
 ; XOPAVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,8,9,2,3,10,11,4,5,12,13,6,7,u,u,16,17,24,25,18,19,26,27,20,21,28,29,22,23,u,u]
+; XOPAVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; XOPAVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; XOPAVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 20, i32 1, i32 21, i32 2, i32 22, i32 3, i32 31, i32 8, i32 28, i32 9, i32 29, i32 10, i32 30, i32 11, i32 31>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_31(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_31:

 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,0,1,4,5,8,9,14,15]
 ; AVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-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:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_04_16_05_17_06_18_07_27_12_24_13_25_14_26_15_27:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [4,16,5,17,6,18,7,27,12,24,13,25,14,26,15,27]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_04_16_05_17_06_18_07_27_12_24_13_25_14_26_15_27:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOPAVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; XOPAVX1-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,u,u,0,1,2,3,4,5],xmm2[6,7]
 ; XOPAVX1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_04_16_05_17_06_18_07_27_12_24_13_25_14_26_15_27:
 ; XOPAVX2:       # %bb.0:
 ; XOPAVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; XOPAVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
+; XOPAVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; XOPAVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7,8,9,10],ymm1[11],ymm0[12,13,14,15]
 ; XOPAVX2-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]
 ; XOPAVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 16, i32 5, i32 17, i32 6, i32 18, i32 7, i32 27, i32 12, i32 24, i32 13, i32 25, i32 14, i32 26, i32 15, i32 27>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_16_01_17_06_22_07_31_08_24_09_25_14_30_15_31(<16 x i16> %a, <16 x i16> %b) {

 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-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:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[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:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; 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:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
 ; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,2,3,6,7,8,9,12,13,10,11,u,u,16,17,20,21,18,19,22,23,24,25,28,29,26,27,u,u]
+; AVX2-FAST-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; 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:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,1,3,20,22,21,31,8,10,9,11,28,30,29,31]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm3 = xmm3[0,1,4,5,2,3,6,7],xmm2[8,9,12,13,10,11,14,15]
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm1 = xmm1[8,9,12,13,10,11],xmm2[14,15],xmm1[u,u,u,u,u,u,u,u]
 ; XOPAVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; XOPAVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_02_01_03_20_22_21_31_08_10_09_11_28_30_29_31:
 ; XOPAVX2:       # %bb.0:
 ; XOPAVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; XOPAVX2-NEXT:    vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
 ; XOPAVX2-NEXT:    vpshuflw {{.*#+}} ymm0 = ymm0[0,2,1,3,4,5,6,7,8,10,9,11,12,13,14,15]
 ; XOPAVX2-NEXT:    vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,6,5,7,8,9,10,11,12,14,13,15]
+; XOPAVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[2,3,2,3]
 ; XOPAVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7],ymm0[8,9,10,11,12,13,14],ymm1[15]
 ; XOPAVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 2, i32 1, i32 3, i32 20, i32 22, i32 21, i32 31, i32 8, i32 10, i32 9, i32 11, i32 28, i32 30, i32 29, i32 31>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-LABEL: shuffle_v16i16_04_04_03_18_uu_uu_uu_uu_12_12_11_26_uu_uu_uu_uu:

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
 ; AVX1-NEXT:    vpshufb {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,10,11,8,9,10,11,12,13,6,7]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-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 {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; AVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; AVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7,8,9,10],ymm0[11],ymm1[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
 ;
 ; AVX512VL-LABEL: shuffle_v16i16_00_01_02_21_20_21_22_11_08_09_10_29_28_29_30_11:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,2,21,20,21,22,11,8,9,10,29,28,29,30,11]
 ; AVX512VL-NEXT:    vpermt2w %ymm1, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
 ;
 ; XOPAVX1-LABEL: shuffle_v16i16_00_01_02_21_20_21_22_11_08_09_10_29_28_29_30_11:
 ; XOPAVX1:       # %bb.0:
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
 ; XOPAVX1-NEXT:    vpperm {{.*#+}} xmm2 = xmm3[0,1,2,3,4,5],xmm2[10,11,8,9,10,11,12,13],xmm3[6,7]
 ; XOPAVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
 ; XOPAVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; XOPAVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3,4,5,6],xmm0[7]
 ; XOPAVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; XOPAVX1-NEXT:    retq
 ;
 ; XOPAVX2-LABEL: shuffle_v16i16_00_01_02_21_20_21_22_11_08_09_10_29_28_29_30_11:
 ; XOPAVX2:       # %bb.0:
 ; XOPAVX2-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
-; XOPAVX2-NEXT:    vperm2i128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; XOPAVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[2,3,2,3]
 ; XOPAVX2-NEXT:    vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7,8,9,10],ymm0[11],ymm1[12,13,14,15]
 ; XOPAVX2-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]
 ; XOPAVX2-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 21, i32 20, i32 21, i32 22, i32 11, i32 8, i32 9, i32 10, i32 29, i32 28, i32 29, i32 30, i32 11>
   ret <16 x i16> %shuffle
 }
 
 define <16 x i16> @shuffle_v16i16_00_17_02_03_20_21_22_15_08_25_10_11_28_29_30_15(<16 x i16> %a, <16 x i16> %b) {

 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-NEXT:    vandps %ymm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: PR34369:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX2-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,10,11,u,u,u,u,u,u,4,5]
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,0,1,0,1,u,u,10,11,4,5,4,5,u,u,30,31,16,17,28,29,16,17,18,19,20,21,24,25,24,25]
-; AVX2-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3],xmm0[4,5,6],xmm2[7]
-; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqw %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpand %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    retq
+; AVX2-SLOW-LABEL: PR34369:
+; AVX2-SLOW:       # %bb.0:
+; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm2 = ymm0[6,7,0,1,0,1,u,u,10,11,4,5,4,5,u,u,30,31,16,17,28,29,16,17,18,19,20,21,24,25,24,25]
+; AVX2-SLOW-NEXT:    vextracti128 $1, %ymm0, %xmm0
+; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,1]
+; AVX2-SLOW-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
+; AVX2-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm2[0,1,2],xmm0[3],xmm2[4,5,6],xmm0[7]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-SLOW-NEXT:    vpcmpeqw %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT:    vpand %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT:    retq
+;
+; AVX2-FAST-LABEL: PR34369:
+; AVX2-FAST:       # %bb.0:
+; AVX2-FAST-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,10,11,u,u,u,u,u,u,4,5]
+; AVX2-FAST-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[6,7,0,1,0,1,u,u,10,11,4,5,4,5,u,u,30,31,16,17,28,29,16,17,18,19,20,21,24,25,24,25]
+; AVX2-FAST-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3],xmm0[4,5,6],xmm2[7]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-FAST-NEXT:    vpcmpeqw %ymm2, %ymm1, %ymm1
+; AVX2-FAST-NEXT:    vpand %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512VL-LABEL: PR34369:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} ymm2 = [3,0,0,13,5,2,2,10,15,8,14,8,9,10,12,12]
 ; AVX512VL-NEXT:    vptestnmw %ymm1, %ymm1, %k1
 ; AVX512VL-NEXT:    vpermw %ymm0, %ymm2, %ymm0 {%k1} {z}
 ; AVX512VL-NEXT:    retq
 ;

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

   ret <8 x float> %shuffle
 }
 
 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:    vperm2f128 {{.*#+}} ymm1 = ymm0[0,1,0,1]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
+; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,0],ymm1[4,4],ymm0[6,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8f32_00040000:
 ; AVX2OR512VL:       # %bb.0:
 ; AVX2OR512VL-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,4,0,0,0,0]
 ; AVX2OR512VL-NEXT:    vpermps %ymm0, %ymm1, %ymm0
 ; AVX2OR512VL-NEXT:    retq

 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x float> %a, <8 x float> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
   ret <8 x float> %shuffle
 }
 
 define <8 x float> @shuffle_v8f32_08991abb(<8 x float> %a, <8 x float> %b) {
 ; AVX1-LABEL: shuffle_v8f32_08991abb:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,0],xmm1[0,0]
+; AVX1-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,1]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8f32_08991abb:
 ; AVX2-SLOW:       # %bb.0:

   ret <8 x i32> %shuffle
 }
 
 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:    vperm2f128 {{.*#+}} ymm1 = ymm0[0,1,0,1]
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4]
+; AVX1-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
 ; AVX1-NEXT:    vshufps {{.*#+}} ymm0 = ymm1[0,0],ymm0[2,0],ymm1[4,4],ymm0[6,4]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2OR512VL-LABEL: shuffle_v8i32_00040000:
 ; AVX2OR512VL:       # %bb.0:
 ; AVX2OR512VL-NEXT:    vmovaps {{.*#+}} ymm1 = [0,0,0,4,0,0,0,0]
 ; AVX2OR512VL-NEXT:    vpermps %ymm0, %ymm1, %ymm0
 ; AVX2OR512VL-NEXT:    retq

 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
   ret <8 x i32> %shuffle
 }
 
 define <8 x i32> @shuffle_v8i32_08991abb(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_08991abb:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,0],xmm1[0,0]
+; AVX1-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,1]
 ; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
 ; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,2,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: shuffle_v8i32_08991abb:
 ; AVX2-SLOW:       # %bb.0:

 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 8, i32 9, i32 9, i32 1, i32 10, i32 11, i32 11>
   ret <8 x i32> %shuffle
 }
 
 define <8 x i32> @shuffle_v8i32_091b2d3f(<8 x i32> %a, <8 x i32> %b) {
 ; AVX1-LABEL: shuffle_v8i32_091b2d3f:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v8i32_091b2d3f:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]

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

 ;
 ; SSE41-LABEL: combine_nested_undef_test15:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,0,1]
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
 ; SSE41-NEXT:    retq
 ;
-; AVX-LABEL: combine_nested_undef_test15:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
-; AVX-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
-; AVX-NEXT:    retq
+; AVX1-LABEL: combine_nested_undef_test15:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
+; AVX1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_nested_undef_test15:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vbroadcastss %xmm1, %xmm1
+; AVX2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,1,0,1]
+; AVX2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
+; AVX2-NEXT:    retq
   %1 = shufflevector <4 x i32> %A, <4 x i32> %B, <4 x i32> <i32 0, i32 4, i32 3, i32 1>
   %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 1, i32 0, i32 3>
   ret <4 x i32> %2
 }
 
 define <4 x i32> @combine_nested_undef_test16(<4 x i32> %A, <4 x i32> %B) {
 ; SSE2-LABEL: combine_nested_undef_test16:
 ; SSE2:       # %bb.0:

   %c = shufflevector <4 x float> %a, <4 x float> %b, <4 x i32><i32 0, i32 4, i32 2, i32 5>
   %d = shufflevector <4 x float> %a, <4 x float> %c, <4 x i32><i32 4, i32 1, i32 5, i32 3>
   ret <4 x float> %d
 }
 
 define <4 x float> @combine_insertps4(<4 x float> %a, <4 x float> %b) {
 ; SSE2-LABEL: combine_insertps4:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: combine_insertps4:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[2,0]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: combine_insertps4:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    insertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
 ; SSE41-NEXT:    retq
 ;

   %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>
   ret <8 x float> %s2
 }
 
 define <4 x float> @PR30264(<4 x float> %x) {
 ; SSE2-LABEL: PR30264:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],mem[2,3]
 ; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: PR30264:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    xorps %xmm1, %xmm1
-; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSSE3-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
 ; SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],mem[2,3]
 ; SSSE3-NEXT:    movaps %xmm1, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: PR30264:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movaps {{.*#+}} xmm1 = <u,u,4.0E+0,1.0E+0>
 ; SSE41-NEXT:    insertps {{.*#+}} xmm1 = xmm0[0],zero,xmm1[2,3]

 ; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_scalar_to_vector_extract:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; SSSE3-NEXT:    psraw $8, %xmm1
-; SSSE3-NEXT:    pextrw $7, %xmm1, %eax
-; SSSE3-NEXT:    movd %eax, %xmm2
 ; SSSE3-NEXT:    movsbl (%rsi), %eax
-; SSSE3-NEXT:    movd %eax, %xmm0
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    movd %eax, %xmm2
+; SSSE3-NEXT:    palignr {{.*#+}} xmm2 = xmm1[14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
 ; SSSE3-NEXT:    movsbl (%rdx), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm0
 ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSSE3-NEXT:    pxor %xmm0, %xmm0
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSSE3-NEXT:    retq

llvm/test/CodeGen/X86/vector-shuffle-variable-128.ll

   ret <16 x i8> %r15
 }
 
 ;
 ; Binary shuffle indices from registers
 ;
 
 define <4 x float> @var_shuffle_v4f32_v4f32_x0yx_i32(<4 x float> %x, <4 x float> %y, i32 %i0, i32 %i1, i32 %i2, i32 %i3) nounwind {
-; SSE-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:
-; SSE:       # %bb.0:
-; SSE-NEXT:    # kill: def $ecx killed $ecx def $rcx
-; SSE-NEXT:    # kill: def $edx killed $edx def $rdx
-; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
-; SSE-NEXT:    andl $3, %edi
-; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    andl $3, %edx
-; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    andl $3, %ecx
-; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; SSE-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE-NEXT:    retq
+; SSE2-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    # kill: def $ecx killed $ecx def $rcx
+; SSE2-NEXT:    # kill: def $edx killed $edx def $rdx
+; SSE2-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSE2-NEXT:    andl $3, %edi
+; SSE2-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE2-NEXT:    andl $3, %edx
+; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE2-NEXT:    andl $3, %ecx
+; SSE2-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSE2-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSE2-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SSE2-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    # kill: def $ecx killed $ecx def $rcx
+; SSSE3-NEXT:    # kill: def $edx killed $edx def $rdx
+; SSSE3-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSSE3-NEXT:    andl $3, %edi
+; SSSE3-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSSE3-NEXT:    andl $3, %edx
+; SSSE3-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSSE3-NEXT:    andl $3, %ecx
+; SSSE3-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSSE3-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSSE3-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SSSE3-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSSE3-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    # kill: def $ecx killed $ecx def $rcx
+; SSE41-NEXT:    # kill: def $edx killed $edx def $rdx
+; SSE41-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSE41-NEXT:    andl $3, %edi
+; SSE41-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE41-NEXT:    andl $3, %edx
+; SSE41-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE41-NEXT:    andl $3, %ecx
+; SSE41-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SSE41-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSE41-NEXT:    insertps {{.*#+}} xmm1 = xmm1[0],mem[0],zero,zero
+; SSE41-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    # kill: def $ecx killed $ecx def $rcx
 ; AVX-NEXT:    # kill: def $edx killed $edx def $rdx
 ; AVX-NEXT:    # kill: def $edi killed $edi def $rdi
 ; AVX-NEXT:    andl $3, %edi
 ; AVX-NEXT:    vmovaps %xmm1, -{{[0-9]+}}(%rsp)
 ; AVX-NEXT:    andl $3, %edx
 ; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
 ; AVX-NEXT:    andl $3, %ecx
 ; AVX-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; AVX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
-; AVX-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
-; AVX-NEXT:    vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; AVX-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],zero,zero
 ; AVX-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX-NEXT:    retq
   %x0 = extractelement <4 x float> %x, i32 %i0
   %x1 = extractelement <4 x float> %x, i32 %i1
   %y2 = extractelement <4 x float> %y, i32 %i2
   %x3 = extractelement <4 x float> %x, i32 %i3
   %r0 = insertelement <4 x float> undef, float %x0, i32 0
   %r1 = insertelement <4 x float>   %r0, float 0.0, i32 1

 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE2-NEXT:    movzwl -24(%rsp,%r9,2), %eax
 ; SSE2-NEXT:    movd %eax, %xmm1
 ; SSE2-NEXT:    movzwl -40(%rsp,%r8,2), %eax
 ; SSE2-NEXT:    movd %eax, %xmm2
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE2-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: var_shuffle_v8i16_v8i16_xyxyxy00_i16:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    # kill: def $r9d killed $r9d def $r9
 ; SSSE3-NEXT:    # kill: def $r8d killed $r8d def $r8
 ; SSSE3-NEXT:    # kill: def $ecx killed $ecx def $rcx
 ; SSSE3-NEXT:    # kill: def $edx killed $edx def $rdx

 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSSE3-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSSE3-NEXT:    movzwl -24(%rsp,%r9,2), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm1
 ; SSSE3-NEXT:    movzwl -40(%rsp,%r8,2), %eax
 ; SSSE3-NEXT:    movd %eax, %xmm2
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; SSSE3-NEXT:    pxor %xmm1, %xmm1
-; SSSE3-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSSE3-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: var_shuffle_v8i16_v8i16_xyxyxy00_i16:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    # kill: def $r9d killed $r9d def $r9
 ; SSE41-NEXT:    # kill: def $r8d killed $r8d def $r8
 ; SSE41-NEXT:    # kill: def $ecx killed $ecx def $rcx
 ; SSE41-NEXT:    # kill: def $edx killed $edx def $rdx

llvm/test/CodeGen/X86/vector-trunc-math.ll

 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_add_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_add_v8i32_v8i16:
 ; AVX2:       # %bb.0:

 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    psraw $8, %xmm0
 ; SSE-NEXT:    paddw %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_add_v8i32_v8i16_sext_8i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
 ; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm0
 ; AVX1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    paddw {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_add_const_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpaddw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_add_const_v8i32_v8i16:

 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_sub_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_sub_v8i32_v8i16:
 ; AVX2:       # %bb.0:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    psubw {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_sub_const_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpsubw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_sub_const_v8i32_v8i16:

 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_mul_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpmulld %xmm1, %xmm0, %xmm2
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpmulld %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_mul_v8i32_v8i16:
 ; AVX2:       # %bb.0:

 ; SSE-NEXT:    psrad $16, %xmm1
 ; SSE-NEXT:    packssdw %xmm2, %xmm1
 ; SSE-NEXT:    pmullw %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_mul_v8i32_v8i16_zext_8i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; AVX1-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    pmullw {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_mul_const_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_mul_const_v8i32_v8i16:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_and_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_and_v8i32_v8i16:
 ; AVX2:       # %bb.0:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    pand {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_and_const_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_and_const_v8i32_v8i16:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_xor_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vxorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_xor_v8i32_v8i16:
 ; AVX2:       # %bb.0:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    pxor {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_xor_const_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpxor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_xor_const_v8i32_v8i16:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_or_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_or_v8i32_v8i16:
 ; AVX2:       # %bb.0:

 ; SSE-NEXT:    psrad $16, %xmm0
 ; SSE-NEXT:    packssdw %xmm1, %xmm0
 ; SSE-NEXT:    por {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_or_const_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_or_const_v8i32_v8i16:

llvm/test/CodeGen/X86/vector-trunc.ll

 ; AVX2-SLOW-NEXT:    vpsrlq $32, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
 ; AVX2-SLOW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,2],ymm2[0,2],ymm0[4,6],ymm2[4,6]
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: trunc8i64_8i32_lshr:
 ; AVX2-FAST:       # %bb.0: # %entry
-; AVX2-FAST-NEXT:    vpsrlq $32, %ymm1, %ymm1
-; AVX2-FAST-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,2,4,6,4,6,6,7]
-; AVX2-FAST-NEXT:    vpermd %ymm0, %ymm2, %ymm0
-; AVX2-FAST-NEXT:    vpermd %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT:    vmovaps {{.*#+}} xmm2 = [1,3,5,7]
+; AVX2-FAST-NEXT:    vpermps %ymm0, %ymm2, %ymm0
+; AVX2-FAST-NEXT:    vpermps %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-FAST-NEXT:    retq
 ;
 ; AVX512-LABEL: trunc8i64_8i32_lshr:
 ; AVX512:       # %bb.0: # %entry
 ; AVX512-NEXT:    vpsrlq $32, %zmm0, %zmm0
 ; AVX512-NEXT:    vpmovqd %zmm0, %ymm0
 ; AVX512-NEXT:    retq
 entry:

 ; SSE41-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3],xmm1[4],xmm2[5],xmm1[6],xmm2[7]
 ; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
 ; SSE41-NEXT:    packusdw %xmm1, %xmm0
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc8i32_8i16:
 ; AVX1:       # %bb.0: # %entry
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc8i32_8i16:
 ; AVX2:       # %bb.0: # %entry

 ; SSE41-NEXT:    shlq $4, %rdi
 ; SSE41-NEXT:    movdqu %xmm0, (%rsi,%rdi)
 ; SSE41-NEXT:    movdqu %xmm2, 16(%rsi,%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: store_merge_split:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,1,4,5,8,9,12,13,u,u,u,u,u,u,u,u>
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm4, %xmm0, %xmm0
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
 ; AVX1-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpshufb %xmm4, %xmm1, %xmm1
 ; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
 ; AVX1-NEXT:    shlq $4, %rdi
 ; AVX1-NEXT:    vmovdqu %xmm0, (%rsi,%rdi)
 ; AVX1-NEXT:    vmovdqu %xmm1, 16(%rsi,%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: store_merge_split:

llvm/test/CodeGen/X86/vector-zext.ll

   %Z = bitcast <16 x i8> %B to <2 x i64>
   ret <2 x i64> %Z
 }
 
 define <4 x i64> @shuf_zext_16i8_to_4i64_offset11(<16 x i8> %A) nounwind uwtable readnone ssp {
 ; SSE2-LABEL: shuf_zext_16i8_to_4i64_offset11:
 ; SSE2:       # %bb.0: # %entry
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero
+; SSE2-NEXT:    psrlq $8, %xmm1
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm0
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSE2-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
 ; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; SSE2-NEXT:    retq

llvm/test/CodeGen/X86/vselect.ll

 ; SSE2-LABEL: simplify_select:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE2-NEXT:    pslld $31, %xmm0
 ; SSE2-NEXT:    psrad $31, %xmm0
 ; SSE2-NEXT:    movd %edi, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,0,1,1]
 ; SSE2-NEXT:    por %xmm1, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm2[1,1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[1,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[1,1]
 ; SSE2-NEXT:    pand %xmm0, %xmm2
 ; SSE2-NEXT:    pandn %xmm1, %xmm0
 ; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: simplify_select:
 ; SSE41:       # %bb.0:

llvm/test/CodeGen/X86/vshift-4.ll

entry:
   store <2 x i64> %shl, <2 x i64>* %dst
   ret void
 }
 
 define void @shift2a(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
 ; X32-LABEL: shift2a:
 ; X32:       # %bb.0: # %entry
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; X32-NEXT:    xorps %xmm2, %xmm2
-; X32-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; X32-NEXT:    pslld %xmm2, %xmm0
+; X32-NEXT:    psrlq $32, %xmm1
+; X32-NEXT:    pslld %xmm1, %xmm0
 ; X32-NEXT:    movdqa %xmm0, (%eax)
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: shift2a:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; X64-NEXT:    xorps %xmm2, %xmm2
-; X64-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; X64-NEXT:    pslld %xmm2, %xmm0
+; X64-NEXT:    psrlq $32, %xmm1
+; X64-NEXT:    pslld %xmm1, %xmm0
 ; X64-NEXT:    movdqa %xmm0, (%rdi)
 ; X64-NEXT:    retq
 entry:
   %shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
   %shl = shl <4 x i32> %val, %shamt
   store <4 x i32> %shl, <4 x i32>* %dst
   ret void
 }
 
 define void @shift2b(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
 ; X32-LABEL: shift2b:
 ; X32:       # %bb.0: # %entry
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; X32-NEXT:    xorps %xmm2, %xmm2
-; X32-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; X32-NEXT:    pslld %xmm2, %xmm0
+; X32-NEXT:    psrlq $32, %xmm1
+; X32-NEXT:    pslld %xmm1, %xmm0
 ; X32-NEXT:    movdqa %xmm0, (%eax)
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: shift2b:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; X64-NEXT:    xorps %xmm2, %xmm2
-; X64-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; X64-NEXT:    pslld %xmm2, %xmm0
+; X64-NEXT:    psrlq $32, %xmm1
+; X64-NEXT:    pslld %xmm1, %xmm0
 ; X64-NEXT:    movdqa %xmm0, (%rdi)
 ; X64-NEXT:    retq
 entry:
   %shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 1, i32 1>
   %shl = shl <4 x i32> %val, %shamt
   store <4 x i32> %shl, <4 x i32>* %dst
   ret void
 }
 
 define void @shift2c(<4 x i32> %val, <4 x i32>* %dst, <2 x i32> %amt) nounwind {
 ; X32-LABEL: shift2c:
 ; X32:       # %bb.0: # %entry
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; X32-NEXT:    xorps %xmm2, %xmm2
-; X32-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; X32-NEXT:    pslld %xmm2, %xmm0
+; X32-NEXT:    psrlq $32, %xmm1
+; X32-NEXT:    pslld %xmm1, %xmm0
 ; X32-NEXT:    movdqa %xmm0, (%eax)
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: shift2c:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; X64-NEXT:    xorps %xmm2, %xmm2
-; X64-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; X64-NEXT:    pslld %xmm2, %xmm0
+; X64-NEXT:    psrlq $32, %xmm1
+; X64-NEXT:    pslld %xmm1, %xmm0
 ; X64-NEXT:    movdqa %xmm0, (%rdi)
 ; X64-NEXT:    retq
 entry:
   %shamt = shufflevector <2 x i32> %amt, <2 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
   %shl = shl <4 x i32> %val, %shamt
   store <4 x i32> %shl, <4 x i32>* %dst
   ret void
 }
 
 define void @shift3a(<8 x i16> %val, <8 x i16>* %dst, <8 x i16> %amt) nounwind {
 ; X32-LABEL: shift3a:
 ; X32:       # %bb.0: # %entry
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X32-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; X32-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1]
+; X32-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,6,6]
 ; X32-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; X32-NEXT:    psllw %xmm1, %xmm0
 ; X32-NEXT:    movdqa %xmm0, (%eax)
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: shift3a:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; X64-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1]
+; X64-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,6,6]
 ; X64-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; X64-NEXT:    psllw %xmm1, %xmm0
 ; X64-NEXT:    movdqa %xmm0, (%rdi)
 ; X64-NEXT:    retq
 entry:
   %shamt = shufflevector <8 x i16> %amt, <8 x i16> undef, <8 x i32> <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
   %shl = shl <8 x i16> %val, %shamt
   store <8 x i16> %shl, <8 x i16>* %dst