[X86] Improve SMULO/UMULO codegen for vXi8 vectors.

[craig.topper]

Summary

The default expansion creates a MUL and either a MULHS/MULHU. Each
of those separately expand to sequences that use one or more
PMULLW instructions as well as additional instructions to
extend the types to vXi16. The MULHS/MULHU expansion computes the
whole 16-bit product, but only keeps the high part.

We can improve the lowering of SMULO/UMULO for some cases by using the MULHS/MULHU
expansion, but keep both the high and low parts. And we can use
those parts to calculate the overflow.

For AVX512 we might have vXi1 overflow outputs. We can improve those by using vpcmpeqw to produce a k register if AVX512BW is enabled. This is a little better than truncating the high result to use vpcmpeqb. If we don't have avx512bw we can extend up to v16i32 to use vpcmpeqd to produce a k register.

[RKSimon]

llvm/lib/Target/X86/X86ISelLowering.cpp

27396

Isn't this really ISD::SMUL_LOHI/UMUL_LOHI ? Could the legalizer be taught to try and use as well?

[craig.topper]

llvm/lib/Target/X86/X86ISelLowering.cpp

27396

The legalizer will try to use SMUL_LOHI/UMUL_LOHI, but I think it tries MULHS/MULHU+MUL first. I suppose scalar was the main consideration there where targets don't usually have both so the order of checking them doesn't matter.

Maybe we could implement SMUL_LOHI/UMUL_LOHI for v16i8 and not MULHS/MULHU. Then fix LegalizeVectorOps to expand MULHS/MULHU to SMUL_LOHI/UMUL_LOHI? It doesn't now and would end up scalarizing if it sees a MULHS/MULHU.

[craig.topper]

Rebase on top of D98587. Merge in the ideas from D97650.
Only use the helper function for the UNPCK lowering. Incorporating the AVX512 mask improvements made the interface weird.

[RKSimon]

llvm/lib/Target/X86/X86ISelLowering.cpp

936

(style) keep columns aligned

1339

(style) keep columns aligned

[craig.topper]

Rebase and fix spacing.

[craig.topper]

clang-format

[RKSimon]

LGTM, I still think in the long run we should be trying to make this ISD::SMUL_LOHI/UMUL_LOHI, but if the legalizers aren't up to the job yet this is a good first step.

llvm/lib/Target/X86/X86ISelLowering.cpp

30147

Use the same indentation as the rest.

llvm/lib/Target/X86/X86ISelLowering.cpp

 //===-- X86ISelLowering.cpp - X86 DAG Lowering Implementation -------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
if (!Subtarget.useSoftFloat() && Subtarget.hasSSE2()) {
     setOperationAction(ISD::MUL,                MVT::v2i64, Custom);
     setOperationAction(ISD::MULHU,              MVT::v4i32, Custom);
     setOperationAction(ISD::MULHS,              MVT::v4i32, Custom);
     setOperationAction(ISD::MULHU,              MVT::v16i8, Custom);
     setOperationAction(ISD::MULHS,              MVT::v16i8, Custom);
     setOperationAction(ISD::MULHU,              MVT::v8i16, Legal);
     setOperationAction(ISD::MULHS,              MVT::v8i16, Legal);
     setOperationAction(ISD::MUL,                MVT::v8i16, Legal);
+
+    setOperationAction(ISD::SMULO, MVT::v16i8, Custom);
+    setOperationAction(ISD::UMULO, MVT::v16i8, Custom);
+
     setOperationAction(ISD::FNEG,               MVT::v2f64, Custom);
     setOperationAction(ISD::FABS,               MVT::v2f64, Custom);
     setOperationAction(ISD::FCOPYSIGN,          MVT::v2f64, Custom);
 
     for (auto VT : { MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64 }) {
       setOperationAction(ISD::SMAX, VT, VT == MVT::v8i16 ? Legal : Custom);
       setOperationAction(ISD::SMIN, VT, VT == MVT::v8i16 ? Legal : Custom);
       setOperationAction(ISD::UMAX, VT, VT == MVT::v16i8 ? Legal : Custom);
if (!Subtarget.useSoftFloat() && Subtarget.hasAVX()) {
 
     setOperationAction(ISD::MULHU,     MVT::v8i32,  Custom);
     setOperationAction(ISD::MULHS,     MVT::v8i32,  Custom);
     setOperationAction(ISD::MULHU,     MVT::v16i16, HasInt256 ? Legal : Custom);
     setOperationAction(ISD::MULHS,     MVT::v16i16, HasInt256 ? Legal : Custom);
     setOperationAction(ISD::MULHU,     MVT::v32i8,  Custom);
     setOperationAction(ISD::MULHS,     MVT::v32i8,  Custom);
 
+    setOperationAction(ISD::SMULO, MVT::v32i8, Custom);
+    setOperationAction(ISD::UMULO, MVT::v32i8, Custom);
+
     setOperationAction(ISD::ABS,       MVT::v4i64,  Custom);
     setOperationAction(ISD::SMAX,      MVT::v4i64,  Custom);
     setOperationAction(ISD::UMAX,      MVT::v4i64,  Custom);
     setOperationAction(ISD::SMIN,      MVT::v4i64,  Custom);
     setOperationAction(ISD::UMIN,      MVT::v4i64,  Custom);
 
     setOperationAction(ISD::UADDSAT,   MVT::v32i8,  HasInt256 ? Legal : Custom);
     setOperationAction(ISD::SADDSAT,   MVT::v32i8,  HasInt256 ? Legal : Custom);
if (!Subtarget.useSoftFloat() && Subtarget.useAVX512Regs()) {
 
     setOperationAction(ISD::MULHU, MVT::v16i32, Custom);
     setOperationAction(ISD::MULHS, MVT::v16i32, Custom);
     setOperationAction(ISD::MULHS, MVT::v32i16, HasBWI ? Legal : Custom);
     setOperationAction(ISD::MULHU, MVT::v32i16, HasBWI ? Legal : Custom);
     setOperationAction(ISD::MULHS, MVT::v64i8,  Custom);
     setOperationAction(ISD::MULHU, MVT::v64i8,  Custom);
 
+    setOperationAction(ISD::SMULO, MVT::v64i8, Custom);
+    setOperationAction(ISD::UMULO, MVT::v64i8, Custom);
+
     setOperationAction(ISD::BITREVERSE, MVT::v64i8,  Custom);
 
     for (auto VT : { MVT::v64i8, MVT::v32i16, MVT::v16i32, MVT::v8i64 }) {
       setOperationAction(ISD::SRL,              VT, Custom);
       setOperationAction(ISD::SHL,              VT, Custom);
       setOperationAction(ISD::SRA,              VT, Custom);
       setOperationAction(ISD::SETCC,            VT, Custom);
 
static SDValue LowerMUL(SDValue Op, const X86Subtarget &Subtarget,
   }
 
   SDValue Hi = DAG.getNode(ISD::ADD, dl, VT, AloBhi, AhiBlo);
   Hi = getTargetVShiftByConstNode(X86ISD::VSHLI, dl, VT, Hi, 32, DAG);
 
   return DAG.getNode(ISD::ADD, dl, VT, AloBlo, Hi);
 }
 
+/// Compute the full 16-bit result of multiplying two vXi8 vectors. Using
+/// PMULLW either by extending to a large vector or splitting the vector in half
+/// and using two PMULLWs. The high 8-bits are returned in a vXi8 vector. The
+/// low 8-bits are returned in the SDValue passed to \p Low.
+static SDValue LowerFullMultiplyvXi8(SDValue A, SDValue B, const SDLoc &dl,
+                                     MVT VT, bool IsSigned,
+                                     const X86Subtarget &Subtarget,
+                                     SelectionDAG &DAG,
+                                     SDValue *Low = nullptr) {
+  unsigned NumElts = VT.getVectorNumElements();
+
+  if ((VT == MVT::v16i8 && Subtarget.hasInt256()) ||
+      (VT == MVT::v32i8 && Subtarget.canExtendTo512BW())) {
+    unsigned NumElts = VT.getVectorNumElements();
+    MVT ExVT = MVT::getVectorVT(MVT::i16, NumElts);
+    unsigned ExAVX = IsSigned ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
+    SDValue ExA = DAG.getNode(ExAVX, dl, ExVT, A);
+    SDValue ExB = DAG.getNode(ExAVX, dl, ExVT, B);
+    SDValue Mul = DAG.getNode(ISD::MUL, dl, ExVT, ExA, ExB);
+
+    if (Low)
+      *Low = DAG.getNode(ISD::TRUNCATE, dl, VT, Mul);
+
+    SDValue High =
+        getTargetVShiftByConstNode(X86ISD::VSRLI, dl, ExVT, Mul, 8, DAG);
+    return DAG.getNode(ISD::TRUNCATE, dl, VT, High);
+  }
+  // For vXi8 we will unpack the low and high half of each 128 bit lane to widen
+  // to a vXi16 type. Do the multiplies, shift the results and pack the half
+  // lane results back together.
+
+  MVT ExVT = MVT::getVectorVT(MVT::i16, NumElts / 2);
+
+  static const int PSHUFDMask[] = {8,  9,  10, 11, 12, 13, 14, 15,
+                                   -1, -1, -1, -1, -1, -1, -1, -1};
+
+  // Extract the lo parts and zero/sign extend to i16.
+  // Only use SSE4.1 instructions for signed v16i8 where using unpack requires
+  // shifts to sign extend. Using unpack for unsigned only requires an xor to
+  // create zeros and a copy due to tied registers contraints pre-avx. But using
+  // zero_extend_vector_inreg would require an additional pshufd for the high
+  // part.
+
+  SDValue ALo, AHi;
+  if (IsSigned && VT == MVT::v16i8 && Subtarget.hasSSE41()) {
+    ALo = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, A);
+
+    AHi = DAG.getVectorShuffle(VT, dl, A, A, PSHUFDMask);
+    AHi = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, AHi);
+  } else if (IsSigned) {
+    ALo = DAG.getBitcast(ExVT, getUnpackl(DAG, dl, VT, DAG.getUNDEF(VT), A));
+    AHi = DAG.getBitcast(ExVT, getUnpackh(DAG, dl, VT, DAG.getUNDEF(VT), A));
+
+    ALo = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, ALo, 8, DAG);
+    AHi = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, AHi, 8, DAG);
+  } else {
+    ALo = DAG.getBitcast(
+        ExVT, getUnpackl(DAG, dl, VT, A, DAG.getConstant(0, dl, VT)));
+    AHi = DAG.getBitcast(
+        ExVT, getUnpackh(DAG, dl, VT, A, DAG.getConstant(0, dl, VT)));
+  }
+
+  SDValue BLo, BHi;
+  if (ISD::isBuildVectorOfConstantSDNodes(B.getNode())) {
+    // If the RHS is a constant, manually unpackl/unpackh and extend.
+    SmallVector<SDValue, 16> LoOps, HiOps;
+    for (unsigned i = 0; i != NumElts; i += 16) {
+      for (unsigned j = 0; j != 8; ++j) {
+        SDValue LoOp = B.getOperand(i + j);
+        SDValue HiOp = B.getOperand(i + j + 8);
+
+        if (IsSigned) {
+          LoOp = DAG.getSExtOrTrunc(LoOp, dl, MVT::i16);
+          HiOp = DAG.getSExtOrTrunc(HiOp, dl, MVT::i16);
+        } else {
+          LoOp = DAG.getZExtOrTrunc(LoOp, dl, MVT::i16);
+          HiOp = DAG.getZExtOrTrunc(HiOp, dl, MVT::i16);
+        }
+
+        LoOps.push_back(LoOp);
+        HiOps.push_back(HiOp);
+      }
+    }
+
+    BLo = DAG.getBuildVector(ExVT, dl, LoOps);
+    BHi = DAG.getBuildVector(ExVT, dl, HiOps);
+  } else if (IsSigned && VT == MVT::v16i8 && Subtarget.hasSSE41()) {
+    BLo = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, B);
+
+    BHi = DAG.getVectorShuffle(VT, dl, B, B, PSHUFDMask);
+    BHi = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, BHi);
+  } else if (IsSigned) {
+    BLo = DAG.getBitcast(ExVT, getUnpackl(DAG, dl, VT, DAG.getUNDEF(VT), B));
+    BHi = DAG.getBitcast(ExVT, getUnpackh(DAG, dl, VT, DAG.getUNDEF(VT), B));
+
+    BLo = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, BLo, 8, DAG);
+    BHi = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, BHi, 8, DAG);
+  } else {
+    BLo = DAG.getBitcast(
+        ExVT, getUnpackl(DAG, dl, VT, B, DAG.getConstant(0, dl, VT)));
+    BHi = DAG.getBitcast(
+        ExVT, getUnpackh(DAG, dl, VT, B, DAG.getConstant(0, dl, VT)));
+  }
+
+  // Multiply, lshr the upper 8bits to the lower 8bits of the lo/hi results and
+  // pack back to vXi8.
+  SDValue RLo = DAG.getNode(ISD::MUL, dl, ExVT, ALo, BLo);
+  SDValue RHi = DAG.getNode(ISD::MUL, dl, ExVT, AHi, BHi);
+
+  if (Low) {
+    // Mask the lower bits and pack the results to rejoin the halves.
+    SDValue Mask = DAG.getConstant(255, dl, ExVT);
+    SDValue LLo = DAG.getNode(ISD::AND, dl, ExVT, RLo, Mask);
+    SDValue LHi = DAG.getNode(ISD::AND, dl, ExVT, RHi, Mask);
+    *Low = DAG.getNode(X86ISD::PACKUS, dl, VT, LLo, LHi);
+  }
+
+  // Shift the high bits and pack those to rejoin the halves.
+  RLo = getTargetVShiftByConstNode(X86ISD::VSRLI, dl, ExVT, RLo, 8, DAG);
+  RHi = getTargetVShiftByConstNode(X86ISD::VSRLI, dl, ExVT, RHi, 8, DAG);
+
+  // Bitcast back to VT and then pack all the even elements from Lo and Hi.
+  return DAG.getNode(X86ISD::PACKUS, dl, VT, RLo, RHi);
+}
+
 static SDValue LowerMULH(SDValue Op, const X86Subtarget &Subtarget,
                          SelectionDAG &DAG) {
   SDLoc dl(Op);
   MVT VT = Op.getSimpleValueType();
   bool IsSigned = Op->getOpcode() == ISD::MULHS;
   unsigned NumElts = VT.getVectorNumElements();
   SDValue A = Op.getOperand(0);
   SDValue B = Op.getOperand(1);
static SDValue LowerMULH(SDValue Op, const X86Subtarget &Subtarget,
   // Only i8 vectors should need custom lowering after this.
   assert((VT == MVT::v16i8 || (VT == MVT::v32i8 && Subtarget.hasInt256()) ||
          (VT == MVT::v64i8 && Subtarget.hasBWI())) &&
          "Unsupported vector type");
 
   // Lower v16i8/v32i8 as extension to v8i16/v16i16 vector pairs, multiply,
   // logical shift down the upper half and pack back to i8.
 
-  // With SSE41 we can use sign/zero extend, but for pre-SSE41 we unpack
-  // and then ashr/lshr the upper bits down to the lower bits before multiply.
-  unsigned ExAVX = IsSigned ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
-
-  if ((VT == MVT::v16i8 && Subtarget.hasInt256()) ||
-      (VT == MVT::v32i8 && Subtarget.canExtendTo512BW())) {
-    MVT ExVT = MVT::getVectorVT(MVT::i16, NumElts);
-    SDValue ExA = DAG.getNode(ExAVX, dl, ExVT, A);
-    SDValue ExB = DAG.getNode(ExAVX, dl, ExVT, B);
-    SDValue Mul = DAG.getNode(ISD::MUL, dl, ExVT, ExA, ExB);
-    Mul = getTargetVShiftByConstNode(X86ISD::VSRLI, dl, ExVT, Mul, 8, DAG);
-    return DAG.getNode(ISD::TRUNCATE, dl, VT, Mul);
-  }
+  return LowerFullMultiplyvXi8(A, B, dl, VT, IsSigned, Subtarget, DAG);
 
-  // For vXi8 we will unpack the low and high half of each 128 bit lane to widen
-  // to a vXi16 type. Do the multiplies, shift the results and pack the half
-  // lane results back together.
-
-  MVT ExVT = MVT::getVectorVT(MVT::i16, NumElts / 2);
-
-  static const int PSHUFDMask[] = { 8,  9, 10, 11, 12, 13, 14, 15,
-                                   -1, -1, -1, -1, -1, -1, -1, -1};
+// Custom lowering for SMULO/UMULO.
+static SDValue LowerMULO(SDValue Op, const X86Subtarget &Subtarget,
+                         SelectionDAG &DAG) {
+  SDLoc dl(Op);
+  MVT VT = Op.getSimpleValueType();
+  bool IsSigned = Op->getOpcode() == ISD::SMULO;
+  SDValue A = Op.getOperand(0);
+  SDValue B = Op.getOperand(1);
 
-  // Extract the lo parts and zero/sign extend to i16.
-  // Only use SSE4.1 instructions for signed v16i8 where using unpack requires
-  // shifts to sign extend. Using unpack for unsigned only requires an xor to
-  // create zeros and a copy due to tied registers contraints pre-avx. But using
-  // zero_extend_vector_inreg would require an additional pshufd for the high
-  // part.
+  // Scalars defer to LowerXALUO.
+  if (!VT.isVector())
+    return LowerXALUO(Op, DAG);
 
-  SDValue ALo, AHi;
-  if (IsSigned && VT == MVT::v16i8 && Subtarget.hasSSE41()) {
-    ALo = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, A);
+  if ((VT == MVT::v32i8 && !Subtarget.hasInt256()) ||
+      (VT == MVT::v64i8 && !Subtarget.hasBWI())) {
 
-    AHi = DAG.getVectorShuffle(VT, dl, A, A, PSHUFDMask);
-    AHi = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, AHi);
-  } else if (IsSigned) {
-    ALo = DAG.getBitcast(ExVT, getUnpackl(DAG, dl, VT, DAG.getUNDEF(VT), A));
-    AHi = DAG.getBitcast(ExVT, getUnpackh(DAG, dl, VT, DAG.getUNDEF(VT), A));
+    // Extract the LHS Lo/Hi vectors
+    SDValue LHS1, LHS2;
+    std::tie(LHS1, LHS2) = splitVector(Op.getOperand(0), DAG, dl);
 
-    ALo = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, ALo, 8, DAG);
-    AHi = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, AHi, 8, DAG);
-  } else {
-    ALo = DAG.getBitcast(ExVT, getUnpackl(DAG, dl, VT, A,
-                                          DAG.getConstant(0, dl, VT)));
-    AHi = DAG.getBitcast(ExVT, getUnpackh(DAG, dl, VT, A,
-                                          DAG.getConstant(0, dl, VT)));
-  }
+    // Extract the RHS Lo/Hi vectors
+    SDValue RHS1, RHS2;
+    std::tie(RHS1, RHS2) = splitVector(Op.getOperand(1), DAG, dl);
 
-  SDValue BLo, BHi;
-  if (ISD::isBuildVectorOfConstantSDNodes(B.getNode())) {
-    // If the RHS is a constant, manually unpackl/unpackh and extend.
-    SmallVector<SDValue, 16> LoOps, HiOps;
-    for (unsigned i = 0; i != NumElts; i += 16) {
-      for (unsigned j = 0; j != 8; ++j) {
-        SDValue LoOp = B.getOperand(i + j);
-        SDValue HiOp = B.getOperand(i + j + 8);
+    EVT OvfVT = Op->getValueType(1);
+    EVT LoOvfVT, HiOvfVT;
+    std::tie(LoOvfVT, HiOvfVT) = DAG.GetSplitDestVTs(OvfVT);
+    SDVTList LoVTs = DAG.getVTList(LHS1.getValueType(), LoOvfVT);
+    SDVTList HiVTs = DAG.getVTList(LHS2.getValueType(), HiOvfVT);
+
+    // Issue the split operations.
+    SDValue Lo = DAG.getNode(Op.getOpcode(), dl, LoVTs, LHS1, RHS1);
+    SDValue Hi = DAG.getNode(Op.getOpcode(), dl, HiVTs, LHS2, RHS2);
 
-        if (IsSigned) {
-          LoOp = DAG.getSExtOrTrunc(LoOp, dl, MVT::i16);
-          HiOp = DAG.getSExtOrTrunc(HiOp, dl, MVT::i16);
-        } else {
-          LoOp = DAG.getZExtOrTrunc(LoOp, dl, MVT::i16);
-          HiOp = DAG.getZExtOrTrunc(HiOp, dl, MVT::i16);
-        }
+    // Join the separate data results and the overflow results.
+    SDValue Res = DAG.getNode(ISD::CONCAT_VECTORS, dl, VT, Lo, Hi);
+    SDValue Ovf = DAG.getNode(ISD::CONCAT_VECTORS, dl, OvfVT, Lo.getValue(1),
+                              Hi.getValue(1));
 
-        LoOps.push_back(LoOp);
-        HiOps.push_back(HiOp);
-      }
-    }
+    return DAG.getMergeValues({Res, Ovf}, dl);
 
-    BLo = DAG.getBuildVector(ExVT, dl, LoOps);
-    BHi = DAG.getBuildVector(ExVT, dl, HiOps);
-  } else if (IsSigned && VT == MVT::v16i8 && Subtarget.hasSSE41()) {
-    BLo = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, B);
+  SDValue Low;
+  SDValue High =
+      LowerFullMultiplyvXi8(A, B, dl, VT, IsSigned, Subtarget, DAG, &Low);
 
-    BHi = DAG.getVectorShuffle(VT, dl, B, B, PSHUFDMask);
-    BHi = DAG.getNode(ISD::SIGN_EXTEND_VECTOR_INREG, dl, ExVT, BHi);
-  } else if (IsSigned) {
-    BLo = DAG.getBitcast(ExVT, getUnpackl(DAG, dl, VT, DAG.getUNDEF(VT), B));
-    BHi = DAG.getBitcast(ExVT, getUnpackh(DAG, dl, VT, DAG.getUNDEF(VT), B));
+  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+  EVT SetccVT =
+      TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), VT);
 
-    BLo = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, BLo, 8, DAG);
-    BHi = getTargetVShiftByConstNode(X86ISD::VSRAI, dl, ExVT, BHi, 8, DAG);
+  SDValue Ovf;
+  if (IsSigned) {
+    // SMULO overflows if the high bits doesn't match the sign of the low.
+    SDValue LowSign =
+        DAG.getNode(ISD::SRA, dl, VT, Low, DAG.getConstant(7, dl, VT));
+    Ovf = DAG.getSetCC(dl, SetccVT, LowSign, High, ISD::SETNE);
   } else {
-    BLo = DAG.getBitcast(ExVT, getUnpackl(DAG, dl, VT, B,
-                                          DAG.getConstant(0, dl, VT)));
-    BHi = DAG.getBitcast(ExVT, getUnpackh(DAG, dl, VT, B,
-                                          DAG.getConstant(0, dl, VT)));
+    // UMULO overflow is the high bits are non-zero.
+    Ovf =
+        DAG.getSetCC(dl, SetccVT, High, DAG.getConstant(0, dl, VT), ISD::SETNE);
   }
 
-  // Multiply, lshr the upper 8bits to the lower 8bits of the lo/hi results and
-  // pack back to vXi8.
-  SDValue RLo = DAG.getNode(ISD::MUL, dl, ExVT, ALo, BLo);
-  SDValue RHi = DAG.getNode(ISD::MUL, dl, ExVT, AHi, BHi);
-  RLo = getTargetVShiftByConstNode(X86ISD::VSRLI, dl, ExVT, RLo, 8, DAG);
-  RHi = getTargetVShiftByConstNode(X86ISD::VSRLI, dl, ExVT, RHi, 8, DAG);
+  Ovf = DAG.getSExtOrTrunc(Ovf, dl, Op->getValueType(1));
 
-  // Bitcast back to VT and then pack all the even elements from Lo and Hi.
-  return DAG.getNode(X86ISD::PACKUS, dl, VT, RLo, RHi);
+  return DAG.getMergeValues({Low, Ovf}, dl);
 }
 
 SDValue X86TargetLowering::LowerWin64_i128OP(SDValue Op, SelectionDAG &DAG) const {
   assert(Subtarget.isTargetWin64() && "Unexpected target");
   EVT VT = Op.getValueType();
   assert(VT.isInteger() && VT.getSizeInBits() == 128 &&
          "Unexpected return type for lowering");
 
SDValue X86TargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const {
   case ISD::ROTR:               return LowerRotate(Op, Subtarget, DAG);
   case ISD::SRA:
   case ISD::SRL:
   case ISD::SHL:                return LowerShift(Op, Subtarget, DAG);
   case ISD::SADDO:
   case ISD::UADDO:
   case ISD::SSUBO:
   case ISD::USUBO:
+    return LowerXALUO(Op, DAG);
   case ISD::SMULO:
-  case ISD::UMULO:              return LowerXALUO(Op, DAG);
+  case ISD::UMULO:
+    return LowerMULO(Op, Subtarget, DAG);
   case ISD::READCYCLECOUNTER:   return LowerREADCYCLECOUNTER(Op, Subtarget,DAG);
   case ISD::BITCAST:            return LowerBITCAST(Op, Subtarget, DAG);
   case ISD::SADDO_CARRY:
   case ISD::SSUBO_CARRY:
   case ISD::ADDCARRY:
   case ISD::SUBCARRY:           return LowerADDSUBCARRY(Op, DAG);
   case ISD::ADD:
   case ISD::SUB:                return lowerAddSub(Op, DAG, Subtarget);

llvm/test/CodeGen/X86/vec_smulo.ll

; AVX512-NEXT:    retq
   %res = sext <16 x i1> %obit to <16 x i32>
   store <16 x i32> %val, <16 x i32>* %p2
   ret <16 x i32> %res
 }
 
 define <16 x i32> @smulo_v16i8(<16 x i8> %a0, <16 x i8> %a1, <16 x i8>* %p2) nounwind {
 ; SSE2-LABEL: smulo_v16i8:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3],xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm1, %xmm0
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm2, %xmm4
+; 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:    psraw $8, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
+; SSE2-NEXT:    psraw $8, %xmm3
+; SSE2-NEXT:    pmullw %xmm2, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    psraw $8, %xmm1
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE2-NEXT:    psraw $8, %xmm4
+; SSE2-NEXT:    pmullw %xmm1, %xmm4
+; SSE2-NEXT:    movdqa %xmm4, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    packuswb %xmm2, %xmm0
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm3
 ; SSE2-NEXT:    pand %xmm1, %xmm4
-; SSE2-NEXT:    packuswb %xmm0, %xmm4
-; SSE2-NEXT:    pxor %xmm0, %xmm0
-; SSE2-NEXT:    pcmpgtb %xmm4, %xmm0
-; SSE2-NEXT:    psraw $8, %xmm3
-; SSE2-NEXT:    psraw $8, %xmm6
-; SSE2-NEXT:    pmullw %xmm3, %xmm6
-; SSE2-NEXT:    psrlw $8, %xmm6
-; SSE2-NEXT:    psraw $8, %xmm5
-; SSE2-NEXT:    psraw $8, %xmm7
-; SSE2-NEXT:    pmullw %xmm5, %xmm7
-; SSE2-NEXT:    psrlw $8, %xmm7
-; SSE2-NEXT:    packuswb %xmm6, %xmm7
-; SSE2-NEXT:    pcmpeqb %xmm0, %xmm7
+; SSE2-NEXT:    packuswb %xmm3, %xmm4
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpgtb %xmm4, %xmm1
+; SSE2-NEXT:    pcmpeqb %xmm0, %xmm1
 ; SSE2-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSE2-NEXT:    pxor %xmm7, %xmm3
+; SSE2-NEXT:    pxor %xmm1, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm1
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm0
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm0
 ; SSE2-NEXT:    psrad $31, %xmm0
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm1
 ; SSE2-NEXT:    psrad $31, %xmm1
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE2-NEXT:    movdqa %xmm3, %xmm2
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm2
 ; SSE2-NEXT:    psrad $31, %xmm2
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
 ; SSE2-NEXT:    movdqa %xmm4, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: smulo_v16i8:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    movdqa %xmm1, %xmm2
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm4
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3],xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm1, %xmm0
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm2, %xmm4
+; SSSE3-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]
+; SSSE3-NEXT:    psraw $8, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
+; SSSE3-NEXT:    psraw $8, %xmm3
+; SSSE3-NEXT:    pmullw %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm2
+; SSSE3-NEXT:    psrlw $8, %xmm2
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    psraw $8, %xmm1
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSSE3-NEXT:    psraw $8, %xmm4
+; SSSE3-NEXT:    pmullw %xmm1, %xmm4
+; SSSE3-NEXT:    movdqa %xmm4, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    packuswb %xmm2, %xmm0
 ; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
-; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm3
 ; SSSE3-NEXT:    pand %xmm1, %xmm4
-; SSSE3-NEXT:    packuswb %xmm0, %xmm4
-; SSSE3-NEXT:    pxor %xmm0, %xmm0
-; SSSE3-NEXT:    pcmpgtb %xmm4, %xmm0
-; SSSE3-NEXT:    psraw $8, %xmm3
-; SSSE3-NEXT:    psraw $8, %xmm6
-; SSSE3-NEXT:    pmullw %xmm3, %xmm6
-; SSSE3-NEXT:    psrlw $8, %xmm6
-; SSSE3-NEXT:    psraw $8, %xmm5
-; SSSE3-NEXT:    psraw $8, %xmm7
-; SSSE3-NEXT:    pmullw %xmm5, %xmm7
-; SSSE3-NEXT:    psrlw $8, %xmm7
-; SSSE3-NEXT:    packuswb %xmm6, %xmm7
-; SSSE3-NEXT:    pcmpeqb %xmm0, %xmm7
+; SSSE3-NEXT:    packuswb %xmm3, %xmm4
+; SSSE3-NEXT:    pxor %xmm1, %xmm1
+; SSSE3-NEXT:    pcmpgtb %xmm4, %xmm1
+; SSSE3-NEXT:    pcmpeqb %xmm0, %xmm1
 ; SSSE3-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSSE3-NEXT:    pxor %xmm7, %xmm3
+; SSSE3-NEXT:    pxor %xmm1, %xmm3
 ; SSSE3-NEXT:    movdqa %xmm3, %xmm1
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    movdqa %xmm1, %xmm0
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm0
 ; SSSE3-NEXT:    psrad $31, %xmm0
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm1
 ; SSSE3-NEXT:    psrad $31, %xmm1
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSSE3-NEXT:    movdqa %xmm3, %xmm2
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm2
 ; SSSE3-NEXT:    psrad $31, %xmm2
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
 ; SSSE3-NEXT:    movdqa %xmm4, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: smulo_v16i8:
 ; 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:    pmovsxbw %xmm1, %xmm3
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm4 = 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:    pmovsxbw %xmm0, %xmm6
-; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[2,3,2,3]
-; 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
+; SSE41-NEXT:    pmovsxbw %xmm1, %xmm2
+; SSE41-NEXT:    pmovsxbw %xmm0, %xmm4
 ; SSE41-NEXT:    pmullw %xmm2, %xmm4
+; SSE41-NEXT:    movdqa %xmm4, %xmm2
+; SSE41-NEXT:    psrlw $8, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm1, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm0, %xmm0
+; SSE41-NEXT:    pmullw %xmm1, %xmm0
+; SSE41-NEXT:    movdqa %xmm0, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    packuswb %xmm1, %xmm2
 ; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
-; SSE41-NEXT:    pand %xmm1, %xmm0
 ; SSE41-NEXT:    pand %xmm1, %xmm4
+; SSE41-NEXT:    pand %xmm1, %xmm0
 ; SSE41-NEXT:    packuswb %xmm0, %xmm4
 ; SSE41-NEXT:    pxor %xmm0, %xmm0
 ; SSE41-NEXT:    pcmpgtb %xmm4, %xmm0
-; SSE41-NEXT:    pmullw %xmm3, %xmm6
-; SSE41-NEXT:    psrlw $8, %xmm6
-; SSE41-NEXT:    pmovsxbw %xmm5, %xmm1
-; SSE41-NEXT:    pmovsxbw %xmm7, %xmm2
-; SSE41-NEXT:    pmullw %xmm1, %xmm2
-; SSE41-NEXT:    psrlw $8, %xmm2
-; SSE41-NEXT:    packuswb %xmm2, %xmm6
-; SSE41-NEXT:    pcmpeqb %xmm0, %xmm6
+; SSE41-NEXT:    pcmpeqb %xmm2, %xmm0
 ; SSE41-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSE41-NEXT:    pxor %xmm6, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm3
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm1
 ; SSE41-NEXT:    psrad $31, %xmm1
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm2
 ; SSE41-NEXT:    psrad $31, %xmm2
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm3
 ; SSE41-NEXT:    psrad $31, %xmm3
 ; SSE41-NEXT:    movdqa %xmm4, (%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: smulo_v16i8:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm2
+; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm3
 ; AVX1-NEXT:    vpmullw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT:    vpand %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; 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:    vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpand %xmm3, %xmm4, %xmm3
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtb %xmm2, %xmm3, %xmm3
-; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm4
-; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm5
-; AVX1-NEXT:    vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm0, %xmm4, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm3, %xmm1
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpand %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm2, %xmm2
+; AVX1-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
 ; AVX1-NEXT:    vmovdqa %xmm2, (%rdi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: smulo_v16i8:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpmovsxbw %xmm1, %ymm2
-; AVX2-NEXT:    vpmovsxbw %xmm0, %ymm3
-; AVX2-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vpsrlw $8, %ymm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX2-NEXT:    vpackuswb %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
+; AVX2-NEXT:    vpmovsxbw %xmm1, %ymm1
+; AVX2-NEXT:    vpmovsxbw %xmm0, %ymm0
 ; AVX2-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackuswb %xmm1, %xmm0, %xmm3
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm2
 ; AVX2-NEXT:    vpxor %xmm0, %xmm0, %xmm0
-; AVX2-NEXT:    vpcmpgtb %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpcmpeqb %xmm0, %xmm2, %xmm0
+; AVX2-NEXT:    vpcmpgtb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
 ; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm1
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm1
-; AVX2-NEXT:    vmovdqa %xmm3, (%rdi)
+; AVX2-NEXT:    vmovdqa %xmm2, (%rdi)
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: smulo_v16i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512F-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm2, %xmm2
-; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT:    vpcmpgtb %xmm2, %xmm3, %xmm3
 ; AVX512F-NEXT:    vpmovsxbw %xmm1, %ymm1
 ; AVX512F-NEXT:    vpmovsxbw %xmm0, %ymm0
 ; AVX512F-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm1, %xmm1
+; AVX512F-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512F-NEXT:    vpcmpgtb %xmm1, %xmm2, %xmm2
 ; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} 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,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX512F-NEXT:    vpcmpeqb %xmm0, %xmm2, %xmm0
 ; AVX512F-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
 ; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k1
 ; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512F-NEXT:    vmovdqa %xmm2, (%rdi)
+; AVX512F-NEXT:    vmovdqa %xmm1, (%rdi)
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512BW-LABEL: smulo_v16i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512BW-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
-; AVX512BW-NEXT:    vpmovwb %ymm2, %xmm2
-; AVX512BW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512BW-NEXT:    vpcmpgtb %xmm2, %xmm3, %xmm3
 ; AVX512BW-NEXT:    vpmovsxbw %xmm1, %ymm1
 ; AVX512BW-NEXT:    vpmovsxbw %xmm0, %ymm0
 ; AVX512BW-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
+; AVX512BW-NEXT:    vpmovwb %ymm0, %xmm1
+; AVX512BW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpcmpgtb %xmm1, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpmovwb %ymm0, %xmm0
-; AVX512BW-NEXT:    vpcmpneqb %xmm3, %xmm0, %k1
+; AVX512BW-NEXT:    vpcmpneqb %xmm0, %xmm2, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT:    vmovdqa %xmm2, (%rdi)
+; AVX512BW-NEXT:    vmovdqa %xmm1, (%rdi)
 ; AVX512BW-NEXT:    retq
   %t = call {<16 x i8>, <16 x i1>} @llvm.smul.with.overflow.v16i8(<16 x i8> %a0, <16 x i8> %a1)
   %val = extractvalue {<16 x i8>, <16 x i1>} %t, 0
   %obit = extractvalue {<16 x i8>, <16 x i1>} %t, 1
   %res = sext <16 x i1> %obit to <16 x i32>
   store <16 x i8> %val, <16 x i8>* %p2
   ret <16 x i32> %res
 }
 
 define <32 x i32> @smulo_v32i8(<32 x i8> %a0, <32 x i8> %a1, <32 x i8>* %p2) nounwind {
 ; SSE2-LABEL: smulo_v32i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movq %rdi, %rax
-; SSE2-NEXT:    movdqa %xmm3, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm3[8],xmm7[9],xmm3[9],xmm7[10],xmm3[10],xmm7[11],xmm3[11],xmm7[12],xmm3[12],xmm7[13],xmm3[13],xmm7[14],xmm3[14],xmm7[15],xmm3[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3],xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm9
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; SSE2-NEXT:    psraw $8, %xmm4
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm1[8],xmm6[9],xmm1[9],xmm6[10],xmm1[10],xmm6[11],xmm1[11],xmm6[12],xmm1[12],xmm6[13],xmm1[13],xmm6[14],xmm1[14],xmm6[15],xmm1[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm3, %xmm1
-; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    pand %xmm8, %xmm1
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm10, %xmm9
-; SSE2-NEXT:    pand %xmm8, %xmm9
-; SSE2-NEXT:    packuswb %xmm1, %xmm9
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    pcmpgtb %xmm9, %xmm1
-; SSE2-NEXT:    psraw $8, %xmm7
 ; SSE2-NEXT:    psraw $8, %xmm6
-; SSE2-NEXT:    pmullw %xmm7, %xmm6
-; SSE2-NEXT:    psrlw $8, %xmm6
-; SSE2-NEXT:    psraw $8, %xmm4
-; SSE2-NEXT:    psraw $8, %xmm5
-; SSE2-NEXT:    pmullw %xmm4, %xmm5
-; SSE2-NEXT:    psrlw $8, %xmm5
-; SSE2-NEXT:    packuswb %xmm6, %xmm5
-; SSE2-NEXT:    pcmpeqb %xmm1, %xmm5
-; SSE2-NEXT:    movdqa %xmm2, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3],xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
-; SSE2-NEXT:    movdqa %xmm2, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
-; SSE2-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]
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm7, %xmm0
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm10, %xmm2
-; SSE2-NEXT:    pxor %xmm7, %xmm7
-; SSE2-NEXT:    pand %xmm8, %xmm0
-; SSE2-NEXT:    pand %xmm8, %xmm2
+; SSE2-NEXT:    pmullw %xmm4, %xmm6
+; SSE2-NEXT:    movdqa %xmm6, %xmm4
+; SSE2-NEXT:    psrlw $8, %xmm4
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    psraw $8, %xmm3
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3],xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
+; SSE2-NEXT:    psraw $8, %xmm9
+; SSE2-NEXT:    pmullw %xmm3, %xmm9
+; SSE2-NEXT:    movdqa %xmm9, %xmm7
+; SSE2-NEXT:    psrlw $8, %xmm7
+; SSE2-NEXT:    packuswb %xmm4, %xmm7
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
+; SSE2-NEXT:    pand %xmm1, %xmm6
+; SSE2-NEXT:    pand %xmm1, %xmm9
+; SSE2-NEXT:    packuswb %xmm6, %xmm9
+; SSE2-NEXT:    pxor %xmm3, %xmm3
+; SSE2-NEXT:    pxor %xmm5, %xmm5
+; SSE2-NEXT:    pcmpgtb %xmm9, %xmm5
+; SSE2-NEXT:    pcmpeqb %xmm7, %xmm5
 ; SSE2-NEXT:    pcmpeqd %xmm8, %xmm8
 ; SSE2-NEXT:    pxor %xmm8, %xmm5
-; SSE2-NEXT:    packuswb %xmm0, %xmm2
-; SSE2-NEXT:    pcmpgtb %xmm2, %xmm7
-; SSE2-NEXT:    psraw $8, %xmm4
-; SSE2-NEXT:    psraw $8, %xmm3
-; SSE2-NEXT:    pmullw %xmm4, %xmm3
-; SSE2-NEXT:    psrlw $8, %xmm3
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm2[8],xmm7[9],xmm2[9],xmm7[10],xmm2[10],xmm7[11],xmm2[11],xmm7[12],xmm2[12],xmm7[13],xmm2[13],xmm7[14],xmm2[14],xmm7[15],xmm2[15]
+; SSE2-NEXT:    psraw $8, %xmm7
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
 ; SSE2-NEXT:    psraw $8, %xmm6
-; SSE2-NEXT:    psraw $8, %xmm1
-; SSE2-NEXT:    pmullw %xmm6, %xmm1
-; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    packuswb %xmm3, %xmm1
-; SSE2-NEXT:    pcmpeqb %xmm7, %xmm1
-; SSE2-NEXT:    pxor %xmm8, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    pmullw %xmm7, %xmm6
+; SSE2-NEXT:    movdqa %xmm6, %xmm7
+; SSE2-NEXT:    psrlw $8, %xmm7
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    psraw $8, %xmm2
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm0, %xmm8
+; SSE2-NEXT:    psraw $8, %xmm0
+; SSE2-NEXT:    pmullw %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    packuswb %xmm7, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm6
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    packuswb %xmm6, %xmm0
+; SSE2-NEXT:    pcmpgtb %xmm0, %xmm3
+; SSE2-NEXT:    pcmpeqb %xmm2, %xmm3
+; SSE2-NEXT:    pxor %xmm8, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm2
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    movdqa %xmm2, %xmm8
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm8
 ; SSE2-NEXT:    psrad $31, %xmm8
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm0
-; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pslld $31, %xmm2
+; SSE2-NEXT:    psrad $31, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE2-NEXT:    movdqa %xmm3, %xmm6
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm6
 ; SSE2-NEXT:    psrad $31, %xmm6
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pslld $31, %xmm3
+; SSE2-NEXT:    psrad $31, %xmm3
+; SSE2-NEXT:    movdqa %xmm5, %xmm7
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    movdqa %xmm7, %xmm1
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm1
 ; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    movdqa %xmm5, %xmm3
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm3, %xmm7
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm7
 ; SSE2-NEXT:    psrad $31, %xmm7
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm3
-; SSE2-NEXT:    psrad $31, %xmm3
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE2-NEXT:    movdqa %xmm5, %xmm4
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm4
 ; SSE2-NEXT:    psrad $31, %xmm4
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm5
 ; SSE2-NEXT:    psrad $31, %xmm5
 ; SSE2-NEXT:    movdqa %xmm9, 16(%rsi)
-; SSE2-NEXT:    movdqa %xmm2, (%rsi)
+; SSE2-NEXT:    movdqa %xmm0, (%rsi)
 ; SSE2-NEXT:    movdqa %xmm5, 112(%rdi)
 ; SSE2-NEXT:    movdqa %xmm4, 96(%rdi)
-; SSE2-NEXT:    movdqa %xmm3, 80(%rdi)
-; SSE2-NEXT:    movdqa %xmm7, 64(%rdi)
-; SSE2-NEXT:    movdqa %xmm1, 48(%rdi)
+; SSE2-NEXT:    movdqa %xmm7, 80(%rdi)
+; SSE2-NEXT:    movdqa %xmm1, 64(%rdi)
+; SSE2-NEXT:    movdqa %xmm3, 48(%rdi)
 ; SSE2-NEXT:    movdqa %xmm6, 32(%rdi)
-; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
+; SSE2-NEXT:    movdqa %xmm2, 16(%rdi)
 ; SSE2-NEXT:    movdqa %xmm8, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: smulo_v32i8:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    movq %rdi, %rax
-; SSSE3-NEXT:    movdqa %xmm3, %xmm10
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm3[8],xmm7[9],xmm3[9],xmm7[10],xmm3[10],xmm7[11],xmm3[11],xmm7[12],xmm3[12],xmm7[13],xmm3[13],xmm7[14],xmm3[14],xmm7[15],xmm3[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3],xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm9
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
+; SSSE3-NEXT:    psraw $8, %xmm4
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm1[8],xmm6[9],xmm1[9],xmm6[10],xmm1[10],xmm6[11],xmm1[11],xmm6[12],xmm1[12],xmm6[13],xmm1[13],xmm6[14],xmm1[14],xmm6[15],xmm1[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm3, %xmm1
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSSE3-NEXT:    pand %xmm8, %xmm1
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm10, %xmm9
-; SSSE3-NEXT:    pand %xmm8, %xmm9
-; SSSE3-NEXT:    packuswb %xmm1, %xmm9
-; SSSE3-NEXT:    pxor %xmm1, %xmm1
-; SSSE3-NEXT:    pcmpgtb %xmm9, %xmm1
-; SSSE3-NEXT:    psraw $8, %xmm7
 ; SSSE3-NEXT:    psraw $8, %xmm6
-; SSSE3-NEXT:    pmullw %xmm7, %xmm6
-; SSSE3-NEXT:    psrlw $8, %xmm6
-; SSSE3-NEXT:    psraw $8, %xmm4
-; SSSE3-NEXT:    psraw $8, %xmm5
-; SSSE3-NEXT:    pmullw %xmm4, %xmm5
-; SSSE3-NEXT:    psrlw $8, %xmm5
-; SSSE3-NEXT:    packuswb %xmm6, %xmm5
-; SSSE3-NEXT:    pcmpeqb %xmm1, %xmm5
-; SSSE3-NEXT:    movdqa %xmm2, %xmm10
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3],xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm2
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
-; 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:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm7, %xmm0
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm10, %xmm2
-; SSSE3-NEXT:    pxor %xmm7, %xmm7
-; SSSE3-NEXT:    pand %xmm8, %xmm0
-; SSSE3-NEXT:    pand %xmm8, %xmm2
+; SSSE3-NEXT:    pmullw %xmm4, %xmm6
+; SSSE3-NEXT:    movdqa %xmm6, %xmm4
+; SSSE3-NEXT:    psrlw $8, %xmm4
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    psraw $8, %xmm3
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3],xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
+; SSSE3-NEXT:    psraw $8, %xmm9
+; SSSE3-NEXT:    pmullw %xmm3, %xmm9
+; SSSE3-NEXT:    movdqa %xmm9, %xmm7
+; SSSE3-NEXT:    psrlw $8, %xmm7
+; SSSE3-NEXT:    packuswb %xmm4, %xmm7
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
+; SSSE3-NEXT:    pand %xmm1, %xmm6
+; SSSE3-NEXT:    pand %xmm1, %xmm9
+; SSSE3-NEXT:    packuswb %xmm6, %xmm9
+; SSSE3-NEXT:    pxor %xmm3, %xmm3
+; SSSE3-NEXT:    pxor %xmm5, %xmm5
+; SSSE3-NEXT:    pcmpgtb %xmm9, %xmm5
+; SSSE3-NEXT:    pcmpeqb %xmm7, %xmm5
 ; SSSE3-NEXT:    pcmpeqd %xmm8, %xmm8
 ; SSSE3-NEXT:    pxor %xmm8, %xmm5
-; SSSE3-NEXT:    packuswb %xmm0, %xmm2
-; SSSE3-NEXT:    pcmpgtb %xmm2, %xmm7
-; SSSE3-NEXT:    psraw $8, %xmm4
-; SSSE3-NEXT:    psraw $8, %xmm3
-; SSSE3-NEXT:    pmullw %xmm4, %xmm3
-; SSSE3-NEXT:    psrlw $8, %xmm3
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm2[8],xmm7[9],xmm2[9],xmm7[10],xmm2[10],xmm7[11],xmm2[11],xmm7[12],xmm2[12],xmm7[13],xmm2[13],xmm7[14],xmm2[14],xmm7[15],xmm2[15]
+; SSSE3-NEXT:    psraw $8, %xmm7
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
 ; SSSE3-NEXT:    psraw $8, %xmm6
-; SSSE3-NEXT:    psraw $8, %xmm1
-; SSSE3-NEXT:    pmullw %xmm6, %xmm1
-; SSSE3-NEXT:    psrlw $8, %xmm1
-; SSSE3-NEXT:    packuswb %xmm3, %xmm1
-; SSSE3-NEXT:    pcmpeqb %xmm7, %xmm1
-; SSSE3-NEXT:    pxor %xmm8, %xmm1
-; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    pmullw %xmm7, %xmm6
+; SSSE3-NEXT:    movdqa %xmm6, %xmm7
+; SSSE3-NEXT:    psrlw $8, %xmm7
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    psraw $8, %xmm2
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm8
+; SSSE3-NEXT:    psraw $8, %xmm0
+; SSSE3-NEXT:    pmullw %xmm2, %xmm0
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    psrlw $8, %xmm2
+; SSSE3-NEXT:    packuswb %xmm7, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm6
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    packuswb %xmm6, %xmm0
+; SSSE3-NEXT:    pcmpgtb %xmm0, %xmm3
+; SSSE3-NEXT:    pcmpeqb %xmm2, %xmm3
+; SSSE3-NEXT:    pxor %xmm8, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm2
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    movdqa %xmm2, %xmm8
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm8
 ; SSSE3-NEXT:    psrad $31, %xmm8
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm0
-; SSSE3-NEXT:    psrad $31, %xmm0
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm6
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    pslld $31, %xmm2
+; SSSE3-NEXT:    psrad $31, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm6
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm6
 ; SSSE3-NEXT:    psrad $31, %xmm6
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    pslld $31, %xmm3
+; SSSE3-NEXT:    psrad $31, %xmm3
+; SSSE3-NEXT:    movdqa %xmm5, %xmm7
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    movdqa %xmm7, %xmm1
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm1
 ; SSSE3-NEXT:    psrad $31, %xmm1
-; SSSE3-NEXT:    movdqa %xmm5, %xmm3
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm3, %xmm7
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm7
 ; SSSE3-NEXT:    psrad $31, %xmm7
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm3
-; SSSE3-NEXT:    psrad $31, %xmm3
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSSE3-NEXT:    movdqa %xmm5, %xmm4
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm4
 ; SSSE3-NEXT:    psrad $31, %xmm4
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm5
 ; SSSE3-NEXT:    psrad $31, %xmm5
 ; SSSE3-NEXT:    movdqa %xmm9, 16(%rsi)
-; SSSE3-NEXT:    movdqa %xmm2, (%rsi)
+; SSSE3-NEXT:    movdqa %xmm0, (%rsi)
 ; SSSE3-NEXT:    movdqa %xmm5, 112(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm4, 96(%rdi)
-; SSSE3-NEXT:    movdqa %xmm3, 80(%rdi)
-; SSSE3-NEXT:    movdqa %xmm7, 64(%rdi)
-; SSSE3-NEXT:    movdqa %xmm1, 48(%rdi)
+; SSSE3-NEXT:    movdqa %xmm7, 80(%rdi)
+; SSSE3-NEXT:    movdqa %xmm1, 64(%rdi)
+; SSSE3-NEXT:    movdqa %xmm3, 48(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm6, 32(%rdi)
-; SSSE3-NEXT:    movdqa %xmm0, 16(%rdi)
+; SSSE3-NEXT:    movdqa %xmm2, 16(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm8, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: smulo_v32i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movq %rdi, %rax
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm10 = 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:    pmovsxbw %xmm3, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm9 = 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:    pmovsxbw %xmm1, %xmm4
-; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE41-NEXT:    pmovsxbw %xmm3, %xmm5
+; SSE41-NEXT:    pmovsxbw %xmm1, %xmm9
+; SSE41-NEXT:    pmullw %xmm5, %xmm9
+; SSE41-NEXT:    movdqa %xmm9, %xmm7
+; SSE41-NEXT:    psrlw $8, %xmm7
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm3, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm1, %xmm1
 ; SSE41-NEXT:    pmullw %xmm3, %xmm1
-; SSE41-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSE41-NEXT:    pand %xmm8, %xmm1
-; SSE41-NEXT:    pmullw %xmm10, %xmm9
-; SSE41-NEXT:    pand %xmm8, %xmm9
+; SSE41-NEXT:    movdqa %xmm1, %xmm3
+; SSE41-NEXT:    psrlw $8, %xmm3
+; SSE41-NEXT:    packuswb %xmm3, %xmm7
+; SSE41-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
+; SSE41-NEXT:    pand %xmm6, %xmm9
+; SSE41-NEXT:    pand %xmm6, %xmm1
 ; SSE41-NEXT:    packuswb %xmm1, %xmm9
+; SSE41-NEXT:    pxor %xmm5, %xmm5
 ; SSE41-NEXT:    pxor %xmm1, %xmm1
 ; SSE41-NEXT:    pcmpgtb %xmm9, %xmm1
-; SSE41-NEXT:    pmullw %xmm7, %xmm4
-; SSE41-NEXT:    psrlw $8, %xmm4
-; SSE41-NEXT:    pmovsxbw %xmm5, %xmm3
-; SSE41-NEXT:    pmovsxbw %xmm6, %xmm5
-; SSE41-NEXT:    pmullw %xmm3, %xmm5
-; SSE41-NEXT:    psrlw $8, %xmm5
-; SSE41-NEXT:    packuswb %xmm5, %xmm4
-; SSE41-NEXT:    pcmpeqb %xmm1, %xmm4
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm10 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; SSE41-NEXT:    pmovsxbw %xmm2, %xmm5
-; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,3,2,3]
-; SSE41-NEXT:    movdqa %xmm2, %xmm7
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; 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:    pmovsxbw %xmm0, %xmm2
-; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm7, %xmm0
-; SSE41-NEXT:    pmullw %xmm10, %xmm1
-; SSE41-NEXT:    pxor %xmm7, %xmm7
-; SSE41-NEXT:    pand %xmm8, %xmm0
-; SSE41-NEXT:    pand %xmm8, %xmm1
+; SSE41-NEXT:    pcmpeqb %xmm7, %xmm1
 ; SSE41-NEXT:    pcmpeqd %xmm8, %xmm8
-; SSE41-NEXT:    pxor %xmm8, %xmm4
-; SSE41-NEXT:    packuswb %xmm0, %xmm1
-; SSE41-NEXT:    pcmpgtb %xmm1, %xmm7
-; SSE41-NEXT:    pmullw %xmm5, %xmm2
+; SSE41-NEXT:    pxor %xmm8, %xmm1
+; SSE41-NEXT:    pmovsxbw %xmm2, %xmm7
+; SSE41-NEXT:    pmovsxbw %xmm0, %xmm3
+; SSE41-NEXT:    pmullw %xmm7, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm7
+; SSE41-NEXT:    psrlw $8, %xmm7
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm2, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm0, %xmm0
+; SSE41-NEXT:    pmullw %xmm2, %xmm0
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
 ; SSE41-NEXT:    psrlw $8, %xmm2
-; SSE41-NEXT:    pmovsxbw %xmm6, %xmm0
-; SSE41-NEXT:    pmovsxbw %xmm3, %xmm3
-; SSE41-NEXT:    pmullw %xmm0, %xmm3
-; SSE41-NEXT:    psrlw $8, %xmm3
-; SSE41-NEXT:    packuswb %xmm3, %xmm2
-; SSE41-NEXT:    pcmpeqb %xmm7, %xmm2
-; SSE41-NEXT:    pxor %xmm8, %xmm2
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; SSE41-NEXT:    packuswb %xmm2, %xmm7
+; SSE41-NEXT:    pand %xmm6, %xmm3
+; SSE41-NEXT:    pand %xmm6, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm3
+; SSE41-NEXT:    pcmpgtb %xmm3, %xmm5
+; SSE41-NEXT:    pcmpeqb %xmm7, %xmm5
+; SSE41-NEXT:    pxor %xmm8, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm8 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm8
 ; SSE41-NEXT:    psrad $31, %xmm8
-; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm5 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm5
-; SSE41-NEXT:    psrad $31, %xmm5
-; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[3,3,3,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm6 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm2
+; SSE41-NEXT:    psrad $31, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[3,3,3,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm6 = xmm6[0],zero,zero,zero,xmm6[1],zero,zero,zero,xmm6[2],zero,zero,zero,xmm6[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm6
 ; SSE41-NEXT:    psrad $31, %xmm6
-; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[1,1,1,1]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm7 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm7 = xmm7[0],zero,zero,zero,xmm7[1],zero,zero,zero,xmm7[2],zero,zero,zero,xmm7[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm7
 ; SSE41-NEXT:    psrad $31, %xmm7
-; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm3
-; SSE41-NEXT:    psrad $31, %xmm3
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm2
-; SSE41-NEXT:    psrad $31, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero,xmm4[2],zero,zero,zero,xmm4[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm4
 ; SSE41-NEXT:    psrad $31, %xmm4
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm5 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm5
+; SSE41-NEXT:    psrad $31, %xmm5
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm1
+; SSE41-NEXT:    psrad $31, %xmm1
 ; SSE41-NEXT:    movdqa %xmm9, 16(%rsi)
-; SSE41-NEXT:    movdqa %xmm1, (%rsi)
-; SSE41-NEXT:    movdqa %xmm4, 64(%rdi)
-; SSE41-NEXT:    movdqa %xmm2, (%rdi)
-; SSE41-NEXT:    movdqa %xmm0, 112(%rdi)
-; SSE41-NEXT:    movdqa %xmm3, 96(%rdi)
+; SSE41-NEXT:    movdqa %xmm3, (%rsi)
+; SSE41-NEXT:    movdqa %xmm1, 64(%rdi)
+; SSE41-NEXT:    movdqa %xmm5, (%rdi)
+; SSE41-NEXT:    movdqa %xmm4, 112(%rdi)
+; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
 ; SSE41-NEXT:    movdqa %xmm7, 80(%rdi)
 ; SSE41-NEXT:    movdqa %xmm6, 48(%rdi)
-; SSE41-NEXT:    movdqa %xmm5, 32(%rdi)
+; SSE41-NEXT:    movdqa %xmm2, 32(%rdi)
 ; SSE41-NEXT:    movdqa %xmm8, 16(%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: smulo_v32i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm3, %xmm4, %xmm3
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT:    vpand %xmm6, %xmm3, %xmm3
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = 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:    vpmovzxbw {{.*#+}} xmm7 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; AVX1-NEXT:    vpmullw %xmm4, %xmm7, %xmm4
-; AVX1-NEXT:    vpand %xmm6, %xmm4, %xmm4
-; AVX1-NEXT:    vpackuswb %xmm3, %xmm4, %xmm9
-; AVX1-NEXT:    vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT:    vpcmpgtb %xmm9, %xmm8, %xmm3
-; AVX1-NEXT:    vpmovsxbw %xmm2, %xmm7
-; AVX1-NEXT:    vpmovsxbw %xmm5, %xmm4
-; AVX1-NEXT:    vpmullw %xmm7, %xmm4, %xmm4
-; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
+; AVX1-NEXT:    vpmovsxbw %xmm2, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm4
+; AVX1-NEXT:    vpmovsxbw %xmm4, %xmm5
+; AVX1-NEXT:    vpmullw %xmm3, %xmm5, %xmm3
+; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm5
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm2, %xmm2
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbw %xmm5, %xmm5
-; AVX1-NEXT:    vpmullw %xmm2, %xmm5, %xmm2
-; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm4, %xmm2
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT:    vpxor %xmm4, %xmm2, %xmm3
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm2, %xmm5, %xmm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbw %xmm4, %xmm4
+; AVX1-NEXT:    vpmullw %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpackuswb %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm6, %xmm3, %xmm3
 ; AVX1-NEXT:    vpand %xmm6, %xmm2, %xmm2
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} 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
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm7 = 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 %xmm5, %xmm7, %xmm5
-; AVX1-NEXT:    vpand %xmm6, %xmm5, %xmm5
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm5, %xmm5
-; AVX1-NEXT:    vpcmpgtb %xmm5, %xmm8, %xmm2
-; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm6
-; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm7
-; AVX1-NEXT:    vpmullw %xmm6, %xmm7, %xmm6
-; AVX1-NEXT:    vpsrlw $8, %xmm6, %xmm6
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm3, %xmm4
+; AVX1-NEXT:    vpxor %xmm8, %xmm8, %xmm8
+; AVX1-NEXT:    vpcmpgtb %xmm4, %xmm8, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm5, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
+; AVX1-NEXT:    vpxor %xmm5, %xmm3, %xmm3
+; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm7
+; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm2
+; AVX1-NEXT:    vpmullw %xmm7, %xmm2, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm7
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm0, %xmm6, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm4, %xmm0, %xmm1
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm7, %xmm1
+; AVX1-NEXT:    vpand %xmm6, %xmm2, %xmm2
+; AVX1-NEXT:    vpand %xmm6, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm2, %xmm6
+; AVX1-NEXT:    vpcmpgtb %xmm6, %xmm8, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm5, %xmm0, %xmm1
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm2
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,1,1]
-; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm4
-; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm4
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[1,1,1,1]
+; AVX1-NEXT:    vpmovsxbd %xmm5, %xmm5
+; AVX1-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm5, %xmm5
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm1
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm4
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm5, %ymm1
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm5, %xmm5
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm4, %ymm3
-; AVX1-NEXT:    vmovdqa %xmm9, 16(%rdi)
-; AVX1-NEXT:    vmovdqa %xmm5, (%rdi)
+; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm3
+; AVX1-NEXT:    vmovdqa %xmm4, 16(%rdi)
+; AVX1-NEXT:    vmovdqa %xmm6, (%rdi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: smulo_v32i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX2-NEXT:    vpsraw $8, %ymm2, %ymm2
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpand %ymm3, %ymm2, %ymm2
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm4 = ymm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm5 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpmullw %ymm4, %ymm5, %ymm4
-; AVX2-NEXT:    vpand %ymm3, %ymm4, %ymm3
-; AVX2-NEXT:    vpackuswb %ymm2, %ymm3, %ymm4
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtb %ymm4, %ymm2, %ymm2
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
 ; AVX2-NEXT:    vpsraw $8, %ymm3, %ymm3
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpsraw $8, %ymm5, %ymm5
-; AVX2-NEXT:    vpmullw %ymm3, %ymm5, %ymm3
-; AVX2-NEXT:    vpsrlw $8, %ymm3, %ymm3
+; AVX2-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $8, %ymm2, %ymm3
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsraw $8, %ymm1, %ymm1
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsraw $8, %ymm0, %ymm0
 ; AVX2-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
-; AVX2-NEXT:    vpackuswb %ymm3, %ymm0, %ymm0
-; AVX2-NEXT:    vpcmpeqb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpackuswb %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-NEXT:    vpand %ymm3, %ymm2, %ymm2
+; AVX2-NEXT:    vpand %ymm3, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm4
+; AVX2-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtb %ymm4, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
 ; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm1
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
 ; AVX2-NEXT:    vpmovsxbd %xmm3, %ymm2
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm1
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm3, %ymm3
 ; AVX2-NEXT:    vmovdqa %ymm4, (%rdi)
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: smulo_v32i8:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm4
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm5 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero,xmm4[8],zero,xmm4[9],zero,xmm4[10],zero,xmm4[11],zero,xmm4[12],zero,xmm4[13],zero,xmm4[14],zero,xmm4[15],zero
-; AVX512F-NEXT:    vpmullw %ymm2, %ymm5, %ymm2
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm2, %xmm2
-; AVX512F-NEXT:    vpxor %xmm5, %xmm5, %xmm5
-; AVX512F-NEXT:    vpcmpgtb %xmm2, %xmm5, %xmm6
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX512F-NEXT:    vpmovsxbw %xmm2, %ymm2
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm3
 ; AVX512F-NEXT:    vpmovsxbw %xmm3, %ymm3
-; AVX512F-NEXT:    vpmovsxbw %xmm4, %ymm4
-; AVX512F-NEXT:    vpmullw %ymm3, %ymm4, %ymm3
-; AVX512F-NEXT:    vpsrlw $8, %ymm3, %ymm3
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm3, %xmm3
-; AVX512F-NEXT:    vpcmpeqb %xmm6, %xmm3, %xmm3
-; AVX512F-NEXT:    vpternlogq $15, %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT:    vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT:    vptestmd %zmm3, %zmm3, %k1
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm4 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512F-NEXT:    vpmullw %ymm3, %ymm4, %ymm3
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
+; AVX512F-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm3, %xmm3
-; AVX512F-NEXT:    vpcmpgtb %xmm3, %xmm5, %xmm4
+; AVX512F-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX512F-NEXT:    vpcmpgtb %xmm3, %xmm4, %xmm5
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm2, %xmm2
+; AVX512F-NEXT:    vpcmpeqb %xmm2, %xmm5, %xmm2
+; AVX512F-NEXT:    vpternlogq $15, %xmm2, %xmm2, %xmm2
+; AVX512F-NEXT:    vpmovsxbd %xmm2, %zmm2
+; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k1
 ; AVX512F-NEXT:    vpmovsxbw %xmm1, %ymm1
 ; AVX512F-NEXT:    vpmovsxbw %xmm0, %ymm0
 ; AVX512F-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm1, %xmm2
+; AVX512F-NEXT:    vpcmpgtb %xmm2, %xmm4, %xmm1
 ; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} 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,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT:    vpcmpeqb %xmm4, %xmm0, %xmm0
+; AVX512F-NEXT:    vpcmpeqb %xmm0, %xmm1, %xmm0
 ; AVX512F-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
 ; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k2
 ; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
 ; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
-; AVX512F-NEXT:    vmovdqa %xmm2, 16(%rdi)
-; AVX512F-NEXT:    vmovdqa %xmm3, (%rdi)
+; AVX512F-NEXT:    vmovdqa %xmm3, 16(%rdi)
+; AVX512F-NEXT:    vmovdqa %xmm2, (%rdi)
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512BW-LABEL: smulo_v32i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm2 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero,ymm1[16],zero,ymm1[17],zero,ymm1[18],zero,ymm1[19],zero,ymm1[20],zero,ymm1[21],zero,ymm1[22],zero,ymm1[23],zero,ymm1[24],zero,ymm1[25],zero,ymm1[26],zero,ymm1[27],zero,ymm1[28],zero,ymm1[29],zero,ymm1[30],zero,ymm1[31],zero
-; AVX512BW-NEXT:    vpmovzxbw {{.*#+}} zmm3 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero
-; AVX512BW-NEXT:    vpmullw %zmm2, %zmm3, %zmm2
-; AVX512BW-NEXT:    vpmovwb %zmm2, %ymm2
-; AVX512BW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512BW-NEXT:    vpcmpgtb %ymm2, %ymm3, %ymm3
 ; AVX512BW-NEXT:    vpmovsxbw %ymm1, %zmm1
 ; AVX512BW-NEXT:    vpmovsxbw %ymm0, %zmm0
 ; AVX512BW-NEXT:    vpmullw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm2
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpcmpgtb %ymm2, %ymm1, %ymm1
 ; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmovwb %zmm0, %ymm0
-; AVX512BW-NEXT:    vpcmpneqb %ymm3, %ymm0, %k1
+; AVX512BW-NEXT:    vpcmpneqb %ymm0, %ymm1, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
 ; AVX512BW-NEXT:    vmovdqa %ymm2, (%rdi)
 ; AVX512BW-NEXT:    retq
   %t = call {<32 x i8>, <32 x i1>} @llvm.smul.with.overflow.v32i8(<32 x i8> %a0, <32 x i8> %a1)
   %val = extractvalue {<32 x i8>, <32 x i1>} %t, 0
   %obit = extractvalue {<32 x i8>, <32 x i1>} %t, 1
   %res = sext <32 x i1> %obit to <32 x i32>
   store <32 x i8> %val, <32 x i8>* %p2
   ret <32 x i32> %res
 }
 
 define <64 x i32> @smulo_v64i8(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %p2) nounwind {
 ; SSE2-LABEL: smulo_v64i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movq %rdi, %rax
-; SSE2-NEXT:    movdqa %xmm7, %xmm9
-; SSE2-NEXT:    movdqa %xmm7, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm8
-; SSE2-NEXT:    movdqa %xmm3, %xmm11
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm10, %xmm11
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm7[8],xmm10[9],xmm7[9],xmm10[10],xmm7[10],xmm10[11],xmm7[11],xmm10[12],xmm7[12],xmm10[13],xmm7[13],xmm10[14],xmm7[14],xmm10[15],xmm7[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm9, %xmm8
-; SSE2-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    pand %xmm9, %xmm11
-; SSE2-NEXT:    pand %xmm9, %xmm8
-; SSE2-NEXT:    packuswb %xmm11, %xmm8
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm3[8],xmm11[9],xmm3[9],xmm11[10],xmm3[10],xmm11[11],xmm3[11],xmm11[12],xmm3[12],xmm11[13],xmm3[13],xmm11[14],xmm3[14],xmm11[15],xmm3[15]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm7[8],xmm8[9],xmm7[9],xmm8[10],xmm7[10],xmm8[11],xmm7[11],xmm8[12],xmm7[12],xmm8[13],xmm7[13],xmm8[14],xmm7[14],xmm8[15],xmm7[15]
+; SSE2-NEXT:    psraw $8, %xmm8
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm3[8],xmm9[9],xmm3[9],xmm9[10],xmm3[10],xmm9[11],xmm3[11],xmm9[12],xmm3[12],xmm9[13],xmm3[13],xmm9[14],xmm3[14],xmm9[15],xmm3[15]
+; SSE2-NEXT:    psraw $8, %xmm9
+; SSE2-NEXT:    pmullw %xmm8, %xmm9
+; SSE2-NEXT:    movdqa %xmm9, %xmm8
+; SSE2-NEXT:    psrlw $8, %xmm8
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1],xmm12[2],xmm3[2],xmm12[3],xmm3[3],xmm12[4],xmm3[4],xmm12[5],xmm3[5],xmm12[6],xmm3[6],xmm12[7],xmm3[7]
-; SSE2-NEXT:    psraw $8, %xmm10
-; SSE2-NEXT:    psraw $8, %xmm11
-; SSE2-NEXT:    pmullw %xmm10, %xmm11
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtb %xmm8, %xmm3
-; SSE2-NEXT:    psrlw $8, %xmm11
-; SSE2-NEXT:    psraw $8, %xmm7
-; SSE2-NEXT:    psraw $8, %xmm12
-; SSE2-NEXT:    pmullw %xmm7, %xmm12
-; SSE2-NEXT:    psrlw $8, %xmm12
-; SSE2-NEXT:    packuswb %xmm11, %xmm12
-; SSE2-NEXT:    pcmpeqb %xmm3, %xmm12
-; SSE2-NEXT:    movdqa %xmm6, %xmm10
-; SSE2-NEXT:    movdqa %xmm6, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm7, %xmm3
-; SSE2-NEXT:    movdqa %xmm2, %xmm11
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm10, %xmm11
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSE2-NEXT:    pand %xmm9, %xmm3
-; SSE2-NEXT:    pand %xmm9, %xmm11
-; SSE2-NEXT:    packuswb %xmm3, %xmm11
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    psraw $8, %xmm7
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm3[0],xmm10[1],xmm3[1],xmm10[2],xmm3[2],xmm10[3],xmm3[3],xmm10[4],xmm3[4],xmm10[5],xmm3[5],xmm10[6],xmm3[6],xmm10[7],xmm3[7]
+; SSE2-NEXT:    psraw $8, %xmm10
+; SSE2-NEXT:    pmullw %xmm7, %xmm10
+; SSE2-NEXT:    movdqa %xmm10, %xmm3
+; SSE2-NEXT:    psrlw $8, %xmm3
+; SSE2-NEXT:    packuswb %xmm8, %xmm3
+; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
+; SSE2-NEXT:    pand %xmm8, %xmm9
+; SSE2-NEXT:    pand %xmm8, %xmm10
+; SSE2-NEXT:    packuswb %xmm9, %xmm10
+; SSE2-NEXT:    pxor %xmm9, %xmm9
+; SSE2-NEXT:    pcmpgtb %xmm10, %xmm9
+; SSE2-NEXT:    pcmpeqb %xmm3, %xmm9
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
 ; SSE2-NEXT:    psraw $8, %xmm3
-; SSE2-NEXT:    pmullw %xmm7, %xmm3
-; SSE2-NEXT:    pxor %xmm7, %xmm7
-; SSE2-NEXT:    pcmpgtb %xmm11, %xmm7
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm2[8],xmm7[9],xmm2[9],xmm7[10],xmm2[10],xmm7[11],xmm2[11],xmm7[12],xmm2[12],xmm7[13],xmm2[13],xmm7[14],xmm2[14],xmm7[15],xmm2[15]
+; SSE2-NEXT:    psraw $8, %xmm7
+; SSE2-NEXT:    pmullw %xmm3, %xmm7
+; SSE2-NEXT:    movdqa %xmm7, %xmm3
 ; SSE2-NEXT:    psrlw $8, %xmm3
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    psraw $8, %xmm6
-; SSE2-NEXT:    psraw $8, %xmm2
-; SSE2-NEXT:    pmullw %xmm6, %xmm2
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3],xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
+; SSE2-NEXT:    psraw $8, %xmm11
+; SSE2-NEXT:    pmullw %xmm6, %xmm11
+; SSE2-NEXT:    movdqa %xmm11, %xmm2
 ; SSE2-NEXT:    psrlw $8, %xmm2
 ; SSE2-NEXT:    packuswb %xmm3, %xmm2
-; SSE2-NEXT:    pcmpeqb %xmm7, %xmm2
-; SSE2-NEXT:    movdqa %xmm5, %xmm3
-; SSE2-NEXT:    movdqa %xmm5, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm6, %xmm7
-; SSE2-NEXT:    movdqa %xmm1, %xmm6
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm3, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
-; SSE2-NEXT:    pand %xmm9, %xmm7
-; SSE2-NEXT:    pand %xmm9, %xmm6
-; SSE2-NEXT:    packuswb %xmm7, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm1[8],xmm7[9],xmm1[9],xmm7[10],xmm1[10],xmm7[11],xmm1[11],xmm7[12],xmm1[12],xmm7[13],xmm1[13],xmm7[14],xmm1[14],xmm7[15],xmm1[15]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pand %xmm8, %xmm7
+; SSE2-NEXT:    pand %xmm8, %xmm11
+; SSE2-NEXT:    packuswb %xmm7, %xmm11
+; SSE2-NEXT:    pxor %xmm6, %xmm6
+; SSE2-NEXT:    pcmpgtb %xmm11, %xmm6
+; SSE2-NEXT:    pcmpeqb %xmm2, %xmm6
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSE2-NEXT:    psraw $8, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
 ; SSE2-NEXT:    psraw $8, %xmm3
+; SSE2-NEXT:    pmullw %xmm2, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3],xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7]
 ; SSE2-NEXT:    psraw $8, %xmm7
-; SSE2-NEXT:    pmullw %xmm3, %xmm7
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtb %xmm6, %xmm3
-; SSE2-NEXT:    psrlw $8, %xmm7
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
 ; SSE2-NEXT:    psraw $8, %xmm5
-; SSE2-NEXT:    psraw $8, %xmm1
-; SSE2-NEXT:    pmullw %xmm5, %xmm1
-; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    packuswb %xmm7, %xmm1
-; SSE2-NEXT:    pcmpeqb %xmm3, %xmm1
-; SSE2-NEXT:    movdqa %xmm4, %xmm3
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm5, %xmm7
-; SSE2-NEXT:    movdqa %xmm0, %xmm5
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm3, %xmm5
-; SSE2-NEXT:    pand %xmm9, %xmm7
-; SSE2-NEXT:    pand %xmm9, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
-; SSE2-NEXT:    packuswb %xmm7, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm0[8],xmm7[9],xmm0[9],xmm7[10],xmm0[10],xmm7[11],xmm0[11],xmm7[12],xmm0[12],xmm7[13],xmm0[13],xmm7[14],xmm0[14],xmm7[15],xmm0[15]
+; SSE2-NEXT:    pmullw %xmm7, %xmm5
+; SSE2-NEXT:    movdqa %xmm5, %xmm7
+; SSE2-NEXT:    psrlw $8, %xmm7
+; SSE2-NEXT:    packuswb %xmm2, %xmm7
+; SSE2-NEXT:    pand %xmm8, %xmm3
+; SSE2-NEXT:    pand %xmm8, %xmm5
+; SSE2-NEXT:    packuswb %xmm3, %xmm5
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpgtb %xmm5, %xmm1
+; SSE2-NEXT:    pcmpeqb %xmm7, %xmm1
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE2-NEXT:    psraw $8, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
 ; SSE2-NEXT:    psraw $8, %xmm3
+; SSE2-NEXT:    pmullw %xmm2, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
 ; SSE2-NEXT:    psraw $8, %xmm7
-; SSE2-NEXT:    pmullw %xmm3, %xmm7
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    psraw $8, %xmm3
-; SSE2-NEXT:    psraw $8, %xmm0
-; SSE2-NEXT:    pmullw %xmm3, %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE2-NEXT:    psraw $8, %xmm4
+; SSE2-NEXT:    pmullw %xmm7, %xmm4
+; SSE2-NEXT:    movdqa %xmm4, %xmm7
 ; SSE2-NEXT:    psrlw $8, %xmm7
-; SSE2-NEXT:    psrlw $8, %xmm0
-; SSE2-NEXT:    packuswb %xmm7, %xmm0
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpgtb %xmm5, %xmm3
-; SSE2-NEXT:    pcmpeqb %xmm3, %xmm0
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSE2-NEXT:    pxor %xmm3, %xmm12
-; SSE2-NEXT:    pxor %xmm3, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm1
-; SSE2-NEXT:    pxor %xmm3, %xmm0
-; SSE2-NEXT:    movdqa %xmm8, 48(%rsi)
+; SSE2-NEXT:    packuswb %xmm2, %xmm7
+; SSE2-NEXT:    pand %xmm8, %xmm3
+; SSE2-NEXT:    pand %xmm8, %xmm4
+; SSE2-NEXT:    packuswb %xmm3, %xmm4
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    pcmpgtb %xmm4, %xmm0
+; SSE2-NEXT:    pcmpeqb %xmm7, %xmm0
+; SSE2-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE2-NEXT:    pxor %xmm2, %xmm9
+; SSE2-NEXT:    pxor %xmm2, %xmm6
+; SSE2-NEXT:    pxor %xmm2, %xmm1
+; SSE2-NEXT:    pxor %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm10, 48(%rsi)
 ; SSE2-NEXT:    movdqa %xmm11, 32(%rsi)
-; SSE2-NEXT:    movdqa %xmm6, 16(%rsi)
-; SSE2-NEXT:    movdqa %xmm12, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm5, (%rsi)
-; SSE2-NEXT:    movdqa %xmm12, %xmm4
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0,0,1,1,2,2,3,3]
-; SSE2-NEXT:    pslld $31, %xmm12
-; SSE2-NEXT:    psrad $31, %xmm12
-; SSE2-NEXT:    movdqa %xmm12, 224(%rdi)
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, 240(%rdi)
-; SSE2-NEXT:    movdqa %xmm3, %xmm4
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    movdqa %xmm5, 16(%rsi)
+; SSE2-NEXT:    movdqa %xmm9, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE2-NEXT:    movdqa %xmm4, (%rsi)
+; SSE2-NEXT:    movdqa %xmm9, %xmm3
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pslld $31, %xmm9
+; SSE2-NEXT:    psrad $31, %xmm9
+; SSE2-NEXT:    movdqa %xmm9, 224(%rdi)
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
-; SSE2-NEXT:    movdqa %xmm3, 192(%rdi)
+; SSE2-NEXT:    movdqa %xmm3, 240(%rdi)
 ; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, 208(%rdi)
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm2
 ; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, 160(%rdi)
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, 176(%rdi)
-; SSE2-NEXT:    movdqa %xmm3, %xmm2
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    movdqa %xmm2, 192(%rdi)
+; SSE2-NEXT:    movdqa %xmm6, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
-; SSE2-NEXT:    movdqa %xmm3, 128(%rdi)
-; SSE2-NEXT:    movdqa %xmm1, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    movdqa %xmm3, 208(%rdi)
+; SSE2-NEXT:    movdqa %xmm6, %xmm3
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pslld $31, %xmm6
+; SSE2-NEXT:    psrad $31, %xmm6
+; SSE2-NEXT:    movdqa %xmm6, 160(%rdi)
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pslld $31, %xmm3
+; SSE2-NEXT:    psrad $31, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, 176(%rdi)
+; SSE2-NEXT:    movdqa %xmm2, %xmm3
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm2
 ; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, 144(%rdi)
+; SSE2-NEXT:    movdqa %xmm2, 128(%rdi)
 ; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pslld $31, %xmm3
+; SSE2-NEXT:    psrad $31, %xmm3
+; SSE2-NEXT:    movdqa %xmm3, 144(%rdi)
+; SSE2-NEXT:    movdqa %xmm1, %xmm3
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm1
 ; SSE2-NEXT:    psrad $31, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, 96(%rdi)
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
-; SSE2-NEXT:    movdqa %xmm2, 112(%rdi)
-; SSE2-NEXT:    movdqa %xmm3, %xmm1
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
-; SSE2-NEXT:    movdqa %xmm3, 64(%rdi)
+; SSE2-NEXT:    movdqa %xmm3, 112(%rdi)
+; SSE2-NEXT:    movdqa %xmm2, %xmm1
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pslld $31, %xmm2
+; SSE2-NEXT:    psrad $31, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, 64(%rdi)
 ; SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm1
 ; SSE2-NEXT:    psrad $31, %xmm1
 ; SSE2-NEXT:    movdqa %xmm1, 80(%rdi)
 ; SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]

 ; SSE2-NEXT:    pslld $31, %xmm0
 ; SSE2-NEXT:    psrad $31, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: smulo_v64i8:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    movq %rdi, %rax
-; SSSE3-NEXT:    movdqa %xmm7, %xmm9
-; SSSE3-NEXT:    movdqa %xmm7, %xmm10
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm3, %xmm8
-; SSSE3-NEXT:    movdqa %xmm3, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm10, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm7[8],xmm10[9],xmm7[9],xmm10[10],xmm7[10],xmm10[11],xmm7[11],xmm10[12],xmm7[12],xmm10[13],xmm7[13],xmm10[14],xmm7[14],xmm10[15],xmm7[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm9, %xmm8
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
-; SSSE3-NEXT:    pand %xmm9, %xmm11
-; SSSE3-NEXT:    pand %xmm9, %xmm8
-; SSSE3-NEXT:    packuswb %xmm11, %xmm8
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm3[8],xmm11[9],xmm3[9],xmm11[10],xmm3[10],xmm11[11],xmm3[11],xmm11[12],xmm3[12],xmm11[13],xmm3[13],xmm11[14],xmm3[14],xmm11[15],xmm3[15]
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm7[8],xmm8[9],xmm7[9],xmm8[10],xmm7[10],xmm8[11],xmm7[11],xmm8[12],xmm7[12],xmm8[13],xmm7[13],xmm8[14],xmm7[14],xmm8[15],xmm7[15]
+; SSSE3-NEXT:    psraw $8, %xmm8
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm3[8],xmm9[9],xmm3[9],xmm9[10],xmm3[10],xmm9[11],xmm3[11],xmm9[12],xmm3[12],xmm9[13],xmm3[13],xmm9[14],xmm3[14],xmm9[15],xmm3[15]
+; SSSE3-NEXT:    psraw $8, %xmm9
+; SSSE3-NEXT:    pmullw %xmm8, %xmm9
+; SSSE3-NEXT:    movdqa %xmm9, %xmm8
+; SSSE3-NEXT:    psrlw $8, %xmm8
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1],xmm12[2],xmm3[2],xmm12[3],xmm3[3],xmm12[4],xmm3[4],xmm12[5],xmm3[5],xmm12[6],xmm3[6],xmm12[7],xmm3[7]
-; SSSE3-NEXT:    psraw $8, %xmm10
-; SSSE3-NEXT:    psraw $8, %xmm11
-; SSSE3-NEXT:    pmullw %xmm10, %xmm11
-; SSSE3-NEXT:    pxor %xmm3, %xmm3
-; SSSE3-NEXT:    pcmpgtb %xmm8, %xmm3
-; SSSE3-NEXT:    psrlw $8, %xmm11
-; SSSE3-NEXT:    psraw $8, %xmm7
-; SSSE3-NEXT:    psraw $8, %xmm12
-; SSSE3-NEXT:    pmullw %xmm7, %xmm12
-; SSSE3-NEXT:    psrlw $8, %xmm12
-; SSSE3-NEXT:    packuswb %xmm11, %xmm12
-; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm12
-; SSSE3-NEXT:    movdqa %xmm6, %xmm10
-; SSSE3-NEXT:    movdqa %xmm6, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm3
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm7, %xmm3
-; SSSE3-NEXT:    movdqa %xmm2, %xmm11
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm10, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSSE3-NEXT:    pand %xmm9, %xmm3
-; SSSE3-NEXT:    pand %xmm9, %xmm11
-; SSSE3-NEXT:    packuswb %xmm3, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    psraw $8, %xmm7
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm3[0],xmm10[1],xmm3[1],xmm10[2],xmm3[2],xmm10[3],xmm3[3],xmm10[4],xmm3[4],xmm10[5],xmm3[5],xmm10[6],xmm3[6],xmm10[7],xmm3[7]
+; SSSE3-NEXT:    psraw $8, %xmm10
+; SSSE3-NEXT:    pmullw %xmm7, %xmm10
+; SSSE3-NEXT:    movdqa %xmm10, %xmm3
+; SSSE3-NEXT:    psrlw $8, %xmm3
+; SSSE3-NEXT:    packuswb %xmm8, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
+; SSSE3-NEXT:    pand %xmm8, %xmm9
+; SSSE3-NEXT:    pand %xmm8, %xmm10
+; SSSE3-NEXT:    packuswb %xmm9, %xmm10
+; SSSE3-NEXT:    pxor %xmm9, %xmm9
+; SSSE3-NEXT:    pcmpgtb %xmm10, %xmm9
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm9
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
 ; SSSE3-NEXT:    psraw $8, %xmm3
-; SSSE3-NEXT:    pmullw %xmm7, %xmm3
-; SSSE3-NEXT:    pxor %xmm7, %xmm7
-; SSSE3-NEXT:    pcmpgtb %xmm11, %xmm7
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm2[8],xmm7[9],xmm2[9],xmm7[10],xmm2[10],xmm7[11],xmm2[11],xmm7[12],xmm2[12],xmm7[13],xmm2[13],xmm7[14],xmm2[14],xmm7[15],xmm2[15]
+; SSSE3-NEXT:    psraw $8, %xmm7
+; SSSE3-NEXT:    pmullw %xmm3, %xmm7
+; SSSE3-NEXT:    movdqa %xmm7, %xmm3
 ; SSSE3-NEXT:    psrlw $8, %xmm3
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    psraw $8, %xmm6
-; SSSE3-NEXT:    psraw $8, %xmm2
-; SSSE3-NEXT:    pmullw %xmm6, %xmm2
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3],xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
+; SSSE3-NEXT:    psraw $8, %xmm11
+; SSSE3-NEXT:    pmullw %xmm6, %xmm11
+; SSSE3-NEXT:    movdqa %xmm11, %xmm2
 ; SSSE3-NEXT:    psrlw $8, %xmm2
 ; SSSE3-NEXT:    packuswb %xmm3, %xmm2
-; SSSE3-NEXT:    pcmpeqb %xmm7, %xmm2
-; SSSE3-NEXT:    movdqa %xmm5, %xmm3
-; SSSE3-NEXT:    movdqa %xmm5, %xmm6
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm6, %xmm7
-; SSSE3-NEXT:    movdqa %xmm1, %xmm6
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm3, %xmm6
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
-; SSSE3-NEXT:    pand %xmm9, %xmm7
-; SSSE3-NEXT:    pand %xmm9, %xmm6
-; SSSE3-NEXT:    packuswb %xmm7, %xmm6
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm1[8],xmm7[9],xmm1[9],xmm7[10],xmm1[10],xmm7[11],xmm1[11],xmm7[12],xmm1[12],xmm7[13],xmm1[13],xmm7[14],xmm1[14],xmm7[15],xmm1[15]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    pand %xmm8, %xmm7
+; SSSE3-NEXT:    pand %xmm8, %xmm11
+; SSSE3-NEXT:    packuswb %xmm7, %xmm11
+; SSSE3-NEXT:    pxor %xmm6, %xmm6
+; SSSE3-NEXT:    pcmpgtb %xmm11, %xmm6
+; SSSE3-NEXT:    pcmpeqb %xmm2, %xmm6
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSSE3-NEXT:    psraw $8, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
 ; SSSE3-NEXT:    psraw $8, %xmm3
+; SSSE3-NEXT:    pmullw %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm2
+; SSSE3-NEXT:    psrlw $8, %xmm2
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3],xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7]
 ; SSSE3-NEXT:    psraw $8, %xmm7
-; SSSE3-NEXT:    pmullw %xmm3, %xmm7
-; SSSE3-NEXT:    pxor %xmm3, %xmm3
-; SSSE3-NEXT:    pcmpgtb %xmm6, %xmm3
-; SSSE3-NEXT:    psrlw $8, %xmm7
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
 ; SSSE3-NEXT:    psraw $8, %xmm5
-; SSSE3-NEXT:    psraw $8, %xmm1
-; SSSE3-NEXT:    pmullw %xmm5, %xmm1
-; SSSE3-NEXT:    psrlw $8, %xmm1
-; SSSE3-NEXT:    packuswb %xmm7, %xmm1
-; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm1
-; SSSE3-NEXT:    movdqa %xmm4, %xmm3
-; SSSE3-NEXT:    movdqa %xmm4, %xmm5
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm5, %xmm7
-; SSSE3-NEXT:    movdqa %xmm0, %xmm5
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm3, %xmm5
-; SSSE3-NEXT:    pand %xmm9, %xmm7
-; SSSE3-NEXT:    pand %xmm9, %xmm5
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
-; SSSE3-NEXT:    packuswb %xmm7, %xmm5
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm0[8],xmm7[9],xmm0[9],xmm7[10],xmm0[10],xmm7[11],xmm0[11],xmm7[12],xmm0[12],xmm7[13],xmm0[13],xmm7[14],xmm0[14],xmm7[15],xmm0[15]
+; SSSE3-NEXT:    pmullw %xmm7, %xmm5
+; SSSE3-NEXT:    movdqa %xmm5, %xmm7
+; SSSE3-NEXT:    psrlw $8, %xmm7
+; SSSE3-NEXT:    packuswb %xmm2, %xmm7
+; SSSE3-NEXT:    pand %xmm8, %xmm3
+; SSSE3-NEXT:    pand %xmm8, %xmm5
+; SSSE3-NEXT:    packuswb %xmm3, %xmm5
+; SSSE3-NEXT:    pxor %xmm1, %xmm1
+; SSSE3-NEXT:    pcmpgtb %xmm5, %xmm1
+; SSSE3-NEXT:    pcmpeqb %xmm7, %xmm1
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSSE3-NEXT:    psraw $8, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
 ; SSSE3-NEXT:    psraw $8, %xmm3
+; SSSE3-NEXT:    pmullw %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm2
+; SSSE3-NEXT:    psrlw $8, %xmm2
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
 ; SSSE3-NEXT:    psraw $8, %xmm7
-; SSSE3-NEXT:    pmullw %xmm3, %xmm7
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    psraw $8, %xmm3
-; SSSE3-NEXT:    psraw $8, %xmm0
-; SSSE3-NEXT:    pmullw %xmm3, %xmm0
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSSE3-NEXT:    psraw $8, %xmm4
+; SSSE3-NEXT:    pmullw %xmm7, %xmm4
+; SSSE3-NEXT:    movdqa %xmm4, %xmm7
 ; SSSE3-NEXT:    psrlw $8, %xmm7
-; SSSE3-NEXT:    psrlw $8, %xmm0
-; SSSE3-NEXT:    packuswb %xmm7, %xmm0
-; SSSE3-NEXT:    pxor %xmm3, %xmm3
-; SSSE3-NEXT:    pcmpgtb %xmm5, %xmm3
-; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
-; SSSE3-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSSE3-NEXT:    pxor %xmm3, %xmm12
-; SSSE3-NEXT:    pxor %xmm3, %xmm2
-; SSSE3-NEXT:    pxor %xmm3, %xmm1
-; SSSE3-NEXT:    pxor %xmm3, %xmm0
-; SSSE3-NEXT:    movdqa %xmm8, 48(%rsi)
+; SSSE3-NEXT:    packuswb %xmm2, %xmm7
+; SSSE3-NEXT:    pand %xmm8, %xmm3
+; SSSE3-NEXT:    pand %xmm8, %xmm4
+; SSSE3-NEXT:    packuswb %xmm3, %xmm4
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    pcmpgtb %xmm4, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm7, %xmm0
+; SSSE3-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSSE3-NEXT:    pxor %xmm2, %xmm9
+; SSSE3-NEXT:    pxor %xmm2, %xmm6
+; SSSE3-NEXT:    pxor %xmm2, %xmm1
+; SSSE3-NEXT:    pxor %xmm2, %xmm0
+; SSSE3-NEXT:    movdqa %xmm10, 48(%rsi)
 ; SSSE3-NEXT:    movdqa %xmm11, 32(%rsi)
-; SSSE3-NEXT:    movdqa %xmm6, 16(%rsi)
-; SSSE3-NEXT:    movdqa %xmm12, %xmm3
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm5, (%rsi)
-; SSSE3-NEXT:    movdqa %xmm12, %xmm4
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0,0,1,1,2,2,3,3]
-; SSSE3-NEXT:    pslld $31, %xmm12
-; SSSE3-NEXT:    psrad $31, %xmm12
-; SSSE3-NEXT:    movdqa %xmm12, 224(%rdi)
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm4
-; SSSE3-NEXT:    psrad $31, %xmm4
-; SSSE3-NEXT:    movdqa %xmm4, 240(%rdi)
-; SSSE3-NEXT:    movdqa %xmm3, %xmm4
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    movdqa %xmm5, 16(%rsi)
+; SSSE3-NEXT:    movdqa %xmm9, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSSE3-NEXT:    movdqa %xmm4, (%rsi)
+; SSSE3-NEXT:    movdqa %xmm9, %xmm3
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    pslld $31, %xmm9
+; SSSE3-NEXT:    psrad $31, %xmm9
+; SSSE3-NEXT:    movdqa %xmm9, 224(%rdi)
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
-; SSSE3-NEXT:    movdqa %xmm3, 192(%rdi)
+; SSSE3-NEXT:    movdqa %xmm3, 240(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm2, %xmm3
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm4
-; SSSE3-NEXT:    psrad $31, %xmm4
-; SSSE3-NEXT:    movdqa %xmm4, 208(%rdi)
-; SSSE3-NEXT:    movdqa %xmm2, %xmm4
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm2
 ; SSSE3-NEXT:    psrad $31, %xmm2
-; SSSE3-NEXT:    movdqa %xmm2, 160(%rdi)
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm4
-; SSSE3-NEXT:    psrad $31, %xmm4
-; SSSE3-NEXT:    movdqa %xmm4, 176(%rdi)
-; SSSE3-NEXT:    movdqa %xmm3, %xmm2
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    movdqa %xmm2, 192(%rdi)
+; SSSE3-NEXT:    movdqa %xmm6, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
-; SSSE3-NEXT:    movdqa %xmm3, 128(%rdi)
-; SSSE3-NEXT:    movdqa %xmm1, %xmm3
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    movdqa %xmm3, 208(%rdi)
+; SSSE3-NEXT:    movdqa %xmm6, %xmm3
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    pslld $31, %xmm6
+; SSSE3-NEXT:    psrad $31, %xmm6
+; SSSE3-NEXT:    movdqa %xmm6, 160(%rdi)
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    pslld $31, %xmm3
+; SSSE3-NEXT:    psrad $31, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, 176(%rdi)
+; SSSE3-NEXT:    movdqa %xmm2, %xmm3
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm2
 ; SSSE3-NEXT:    psrad $31, %xmm2
-; SSSE3-NEXT:    movdqa %xmm2, 144(%rdi)
+; SSSE3-NEXT:    movdqa %xmm2, 128(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    pslld $31, %xmm3
+; SSSE3-NEXT:    psrad $31, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, 144(%rdi)
+; SSSE3-NEXT:    movdqa %xmm1, %xmm3
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm1
 ; SSSE3-NEXT:    psrad $31, %xmm1
 ; SSSE3-NEXT:    movdqa %xmm1, 96(%rdi)
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm2
-; SSSE3-NEXT:    psrad $31, %xmm2
-; SSSE3-NEXT:    movdqa %xmm2, 112(%rdi)
-; SSSE3-NEXT:    movdqa %xmm3, %xmm1
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
-; SSSE3-NEXT:    movdqa %xmm3, 64(%rdi)
+; SSSE3-NEXT:    movdqa %xmm3, 112(%rdi)
+; SSSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    pslld $31, %xmm2
+; SSSE3-NEXT:    psrad $31, %xmm2
+; SSSE3-NEXT:    movdqa %xmm2, 64(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm0, %xmm2
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm1
 ; SSSE3-NEXT:    psrad $31, %xmm1
 ; SSSE3-NEXT:    movdqa %xmm1, 80(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm0, %xmm1
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3]

 ; SSSE3-NEXT:    pslld $31, %xmm0
 ; SSSE3-NEXT:    psrad $31, %xmm0
 ; SSSE3-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: smulo_v64i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movq %rdi, %rax
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm8 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero
-; SSE41-NEXT:    movdqa %xmm7, %xmm10
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm9 = 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:    movdqa %xmm3, %xmm11
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm10, %xmm11
-; SSE41-NEXT:    pmovsxbw %xmm7, %xmm12
-; SSE41-NEXT:    pmullw %xmm8, %xmm9
-; SSE41-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
-; SSE41-NEXT:    pand %xmm10, %xmm11
-; SSE41-NEXT:    pand %xmm10, %xmm9
-; SSE41-NEXT:    packuswb %xmm11, %xmm9
+; SSE41-NEXT:    pmovsxbw %xmm7, %xmm9
 ; SSE41-NEXT:    pmovsxbw %xmm3, %xmm8
-; SSE41-NEXT:    pshufd {{.*#+}} xmm11 = xmm7[2,3,2,3]
-; SSE41-NEXT:    pshufd {{.*#+}} xmm13 = xmm3[2,3,2,3]
-; SSE41-NEXT:    pmullw %xmm12, %xmm8
-; SSE41-NEXT:    pxor %xmm7, %xmm7
-; SSE41-NEXT:    pcmpgtb %xmm9, %xmm7
-; SSE41-NEXT:    psrlw $8, %xmm8
-; SSE41-NEXT:    pmovsxbw %xmm11, %xmm11
-; SSE41-NEXT:    pmovsxbw %xmm13, %xmm3
-; SSE41-NEXT:    pmullw %xmm11, %xmm3
-; SSE41-NEXT:    psrlw $8, %xmm3
-; SSE41-NEXT:    packuswb %xmm3, %xmm8
-; SSE41-NEXT:    pcmpeqb %xmm7, %xmm8
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm11 = 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:    movdqa %xmm6, %xmm7
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm2, %xmm3
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm7, %xmm3
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm12 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; SSE41-NEXT:    pmullw %xmm11, %xmm12
-; SSE41-NEXT:    pmovsxbw %xmm6, %xmm7
-; SSE41-NEXT:    pand %xmm10, %xmm3
-; SSE41-NEXT:    pand %xmm10, %xmm12
-; SSE41-NEXT:    packuswb %xmm3, %xmm12
-; SSE41-NEXT:    pmovsxbw %xmm2, %xmm3
-; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
-; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; SSE41-NEXT:    pmullw %xmm9, %xmm8
+; SSE41-NEXT:    movdqa %xmm8, %xmm10
+; SSE41-NEXT:    psrlw $8, %xmm10
+; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm7, %xmm7
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; SSE41-NEXT:    pmovsxbw %xmm3, %xmm3
 ; SSE41-NEXT:    pmullw %xmm7, %xmm3
-; SSE41-NEXT:    pxor %xmm7, %xmm7
-; SSE41-NEXT:    pcmpgtb %xmm12, %xmm7
+; SSE41-NEXT:    movdqa %xmm3, %xmm7
+; SSE41-NEXT:    psrlw $8, %xmm7
+; SSE41-NEXT:    packuswb %xmm7, %xmm10
+; SSE41-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
+; SSE41-NEXT:    pand %xmm9, %xmm8
+; SSE41-NEXT:    pand %xmm9, %xmm3
+; SSE41-NEXT:    packuswb %xmm3, %xmm8
+; SSE41-NEXT:    pxor %xmm11, %xmm11
+; SSE41-NEXT:    pcmpgtb %xmm8, %xmm11
+; SSE41-NEXT:    pcmpeqb %xmm10, %xmm11
+; SSE41-NEXT:    pmovsxbw %xmm6, %xmm3
+; SSE41-NEXT:    pmovsxbw %xmm2, %xmm7
+; SSE41-NEXT:    pmullw %xmm3, %xmm7
+; SSE41-NEXT:    movdqa %xmm7, %xmm3
 ; SSE41-NEXT:    psrlw $8, %xmm3
+; SSE41-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
 ; SSE41-NEXT:    pmovsxbw %xmm6, %xmm6
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; SSE41-NEXT:    pmovsxbw %xmm2, %xmm2
 ; SSE41-NEXT:    pmullw %xmm6, %xmm2
-; SSE41-NEXT:    psrlw $8, %xmm2
-; SSE41-NEXT:    packuswb %xmm2, %xmm3
-; SSE41-NEXT:    pcmpeqb %xmm7, %xmm3
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm2 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; SSE41-NEXT:    movdqa %xmm5, %xmm6
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm1, %xmm7
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm6, %xmm7
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm11 = 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:    pmullw %xmm2, %xmm11
-; SSE41-NEXT:    pmovsxbw %xmm5, %xmm6
-; SSE41-NEXT:    pand %xmm10, %xmm7
-; SSE41-NEXT:    pand %xmm10, %xmm11
-; SSE41-NEXT:    packuswb %xmm7, %xmm11
-; SSE41-NEXT:    pmovsxbw %xmm1, %xmm2
+; SSE41-NEXT:    movdqa %xmm2, %xmm6
+; SSE41-NEXT:    psrlw $8, %xmm6
+; SSE41-NEXT:    packuswb %xmm6, %xmm3
+; SSE41-NEXT:    pand %xmm9, %xmm7
+; SSE41-NEXT:    pand %xmm9, %xmm2
+; SSE41-NEXT:    packuswb %xmm2, %xmm7
+; SSE41-NEXT:    pxor %xmm2, %xmm2
+; SSE41-NEXT:    pcmpgtb %xmm7, %xmm2
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm2
+; SSE41-NEXT:    pmovsxbw %xmm5, %xmm3
+; SSE41-NEXT:    pmovsxbw %xmm1, %xmm6
+; SSE41-NEXT:    pmullw %xmm3, %xmm6
+; SSE41-NEXT:    movdqa %xmm6, %xmm3
+; SSE41-NEXT:    psrlw $8, %xmm3
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE41-NEXT:    pmullw %xmm6, %xmm2
-; SSE41-NEXT:    pxor %xmm6, %xmm6
-; SSE41-NEXT:    pcmpgtb %xmm11, %xmm6
-; SSE41-NEXT:    psrlw $8, %xmm2
 ; SSE41-NEXT:    pmovsxbw %xmm5, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE41-NEXT:    pmovsxbw %xmm1, %xmm1
 ; SSE41-NEXT:    pmullw %xmm5, %xmm1
-; SSE41-NEXT:    psrlw $8, %xmm1
-; SSE41-NEXT:    packuswb %xmm1, %xmm2
-; SSE41-NEXT:    pcmpeqb %xmm6, %xmm2
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm1 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero
-; SSE41-NEXT:    movdqa %xmm4, %xmm5
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm0, %xmm6
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm5, %xmm6
-; SSE41-NEXT:    pmovzxbw {{.*#+}} 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
-; SSE41-NEXT:    pmullw %xmm1, %xmm5
-; SSE41-NEXT:    pand %xmm10, %xmm6
-; SSE41-NEXT:    pand %xmm10, %xmm5
-; SSE41-NEXT:    pmovsxbw %xmm4, %xmm7
-; SSE41-NEXT:    packuswb %xmm6, %xmm5
-; SSE41-NEXT:    pmovsxbw %xmm0, %xmm1
-; SSE41-NEXT:    pmullw %xmm7, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, %xmm5
+; SSE41-NEXT:    psrlw $8, %xmm5
+; SSE41-NEXT:    packuswb %xmm5, %xmm3
+; SSE41-NEXT:    pand %xmm9, %xmm6
+; SSE41-NEXT:    pand %xmm9, %xmm1
+; SSE41-NEXT:    packuswb %xmm1, %xmm6
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pcmpgtb %xmm6, %xmm1
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm1
+; SSE41-NEXT:    pmovsxbw %xmm4, %xmm3
+; SSE41-NEXT:    pmovsxbw %xmm0, %xmm5
+; SSE41-NEXT:    pmullw %xmm3, %xmm5
+; SSE41-NEXT:    movdqa %xmm5, %xmm3
+; SSE41-NEXT:    psrlw $8, %xmm3
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSE41-NEXT:    pmovsxbw %xmm4, %xmm4
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSE41-NEXT:    pmovsxbw %xmm0, %xmm0
 ; SSE41-NEXT:    pmullw %xmm4, %xmm0
-; SSE41-NEXT:    psrlw $8, %xmm1
-; SSE41-NEXT:    psrlw $8, %xmm0
-; SSE41-NEXT:    packuswb %xmm0, %xmm1
+; SSE41-NEXT:    movdqa %xmm0, %xmm4
+; SSE41-NEXT:    psrlw $8, %xmm4
+; SSE41-NEXT:    packuswb %xmm4, %xmm3
+; SSE41-NEXT:    pand %xmm9, %xmm5
+; SSE41-NEXT:    pand %xmm9, %xmm0
+; SSE41-NEXT:    packuswb %xmm0, %xmm5
 ; SSE41-NEXT:    pxor %xmm0, %xmm0
 ; SSE41-NEXT:    pcmpgtb %xmm5, %xmm0
-; SSE41-NEXT:    pcmpeqb %xmm0, %xmm1
-; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
-; SSE41-NEXT:    pxor %xmm0, %xmm8
-; SSE41-NEXT:    pxor %xmm0, %xmm3
-; SSE41-NEXT:    pxor %xmm0, %xmm2
-; SSE41-NEXT:    pxor %xmm0, %xmm1
-; SSE41-NEXT:    movdqa %xmm9, 48(%rsi)
-; SSE41-NEXT:    movdqa %xmm12, 32(%rsi)
-; SSE41-NEXT:    movdqa %xmm11, 16(%rsi)
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqd %xmm3, %xmm3
+; SSE41-NEXT:    pxor %xmm3, %xmm11
+; SSE41-NEXT:    pxor %xmm3, %xmm2
+; SSE41-NEXT:    pxor %xmm3, %xmm1
+; SSE41-NEXT:    pxor %xmm3, %xmm0
+; SSE41-NEXT:    movdqa %xmm8, 48(%rsi)
+; SSE41-NEXT:    movdqa %xmm7, 32(%rsi)
+; SSE41-NEXT:    movdqa %xmm6, 16(%rsi)
 ; SSE41-NEXT:    movdqa %xmm5, (%rsi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm8[0],zero,zero,zero,xmm8[1],zero,zero,zero,xmm8[2],zero,zero,zero,xmm8[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 192(%rdi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 128(%rdi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 64(%rdi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, (%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 224(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 240(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 208(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 160(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 176(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 144(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 112(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 80(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 32(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,3,3,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    movdqa %xmm0, 48(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm11[0],zero,zero,zero,xmm11[1],zero,zero,zero,xmm11[2],zero,zero,zero,xmm11[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 192(%rdi)
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 128(%rdi)
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 64(%rdi)
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, (%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[2,3,2,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 224(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[3,3,3,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 240(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 208(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 160(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[3,3,3,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm3
+; SSE41-NEXT:    psrad $31, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, 176(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm2
+; SSE41-NEXT:    psrad $31, %xmm2
+; SSE41-NEXT:    movdqa %xmm2, 144(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm2
+; SSE41-NEXT:    psrad $31, %xmm2
+; SSE41-NEXT:    movdqa %xmm2, 96(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm2
+; SSE41-NEXT:    psrad $31, %xmm2
+; SSE41-NEXT:    movdqa %xmm2, 112(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm1
+; SSE41-NEXT:    psrad $31, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, 80(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm1
+; SSE41-NEXT:    psrad $31, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, 32(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm1
+; SSE41-NEXT:    psrad $31, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, 48(%rdi)
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: smulo_v64i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    movq %rdi, %rax
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm5
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm4
+; AVX1-NEXT:    vpmovsxbw %xmm4, %xmm5
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm6
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm7 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm4, %xmm7, %xmm4
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT:    vpand %xmm4, %xmm10, %xmm8
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm7 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero
-; AVX1-NEXT:    vpmullw %xmm7, %xmm4, %xmm4
-; AVX1-NEXT:    vpand %xmm4, %xmm10, %xmm4
-; AVX1-NEXT:    vpackuswb %xmm8, %xmm4, %xmm9
-; AVX1-NEXT:    vpxor %xmm11, %xmm11, %xmm11
-; AVX1-NEXT:    vpcmpgtb %xmm9, %xmm11, %xmm8
-; AVX1-NEXT:    vpmovsxbw %xmm5, %xmm4
 ; AVX1-NEXT:    vpmovsxbw %xmm6, %xmm7
-; AVX1-NEXT:    vpmullw %xmm4, %xmm7, %xmm4
-; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbw %xmm5, %xmm5
+; AVX1-NEXT:    vpmullw %xmm5, %xmm7, %xmm7
+; AVX1-NEXT:    vpsrlw $8, %xmm7, %xmm5
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbw %xmm4, %xmm4
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm6, %xmm6
-; AVX1-NEXT:    vpmullw %xmm5, %xmm6, %xmm5
-; AVX1-NEXT:    vpsrlw $8, %xmm5, %xmm5
-; AVX1-NEXT:    vpackuswb %xmm5, %xmm4, %xmm4
-; AVX1-NEXT:    vpcmpeqb %xmm4, %xmm8, %xmm8
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpand %xmm4, %xmm10, %xmm4
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm5 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX1-NEXT:    vpmullw %xmm5, %xmm6, %xmm5
-; AVX1-NEXT:    vpand %xmm5, %xmm10, %xmm5
-; AVX1-NEXT:    vpackuswb %xmm4, %xmm5, %xmm12
-; AVX1-NEXT:    vpcmpgtb %xmm12, %xmm11, %xmm4
-; AVX1-NEXT:    vpmovsxbw %xmm3, %xmm6
-; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm7
-; AVX1-NEXT:    vpmullw %xmm6, %xmm7, %xmm6
-; AVX1-NEXT:    vpsrlw $8, %xmm6, %xmm6
+; AVX1-NEXT:    vpmullw %xmm4, %xmm6, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm6
+; AVX1-NEXT:    vpackuswb %xmm6, %xmm5, %xmm6
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm5, %xmm7, %xmm7
+; AVX1-NEXT:    vpand %xmm5, %xmm4, %xmm4
+; AVX1-NEXT:    vpackuswb %xmm4, %xmm7, %xmm9
+; AVX1-NEXT:    vpxor %xmm10, %xmm10, %xmm10
+; AVX1-NEXT:    vpcmpgtb %xmm9, %xmm10, %xmm4
+; AVX1-NEXT:    vpcmpeqb %xmm6, %xmm4, %xmm8
+; AVX1-NEXT:    vpmovsxbw %xmm3, %xmm4
+; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm6
+; AVX1-NEXT:    vpmullw %xmm4, %xmm6, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm6
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm3, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm1, %xmm1
 ; AVX1-NEXT:    vpmullw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrlw $8, %xmm1, %xmm1
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm6, %xmm1
-; AVX1-NEXT:    vpcmpeqb %xmm4, %xmm1, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm4
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; AVX1-NEXT:    vpsrlw $8, %xmm1, %xmm3
+; AVX1-NEXT:    vpackuswb %xmm3, %xmm6, %xmm3
+; AVX1-NEXT:    vpand %xmm5, %xmm4, %xmm4
+; AVX1-NEXT:    vpand %xmm5, %xmm1, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm4, %xmm11
+; AVX1-NEXT:    vpcmpgtb %xmm11, %xmm10, %xmm4
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm4, %xmm12
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vpmovsxbw %xmm3, %xmm4
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm6
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm7 = xmm6[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm1, %xmm7, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm10, %xmm1
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm7 = 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:    vpmovzxbw {{.*#+}} xmm5 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero
-; AVX1-NEXT:    vpmullw %xmm7, %xmm5, %xmm5
-; AVX1-NEXT:    vpand %xmm5, %xmm10, %xmm5
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm5, %xmm1
-; AVX1-NEXT:    vpmovsxbw %xmm4, %xmm5
-; AVX1-NEXT:    vpmovsxbw %xmm6, %xmm7
-; AVX1-NEXT:    vpmullw %xmm5, %xmm7, %xmm5
-; AVX1-NEXT:    vpcmpgtb %xmm1, %xmm11, %xmm7
-; AVX1-NEXT:    vpsrlw $8, %xmm5, %xmm5
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbw %xmm4, %xmm4
+; AVX1-NEXT:    vpmovsxbw %xmm6, %xmm1
+; AVX1-NEXT:    vpmullw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpsrlw $8, %xmm1, %xmm4
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbw %xmm3, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm6, %xmm6
-; AVX1-NEXT:    vpmullw %xmm4, %xmm6, %xmm4
-; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
-; AVX1-NEXT:    vpackuswb %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpcmpeqb %xmm7, %xmm4, %xmm4
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm5, %xmm6, %xmm5
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = 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:    vpmovzxbw {{.*#+}} xmm7 = 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 %xmm6, %xmm7, %xmm6
-; AVX1-NEXT:    vpand %xmm5, %xmm10, %xmm5
-; AVX1-NEXT:    vpand %xmm6, %xmm10, %xmm6
-; AVX1-NEXT:    vpackuswb %xmm5, %xmm6, %xmm7
-; AVX1-NEXT:    vpmovsxbw %xmm2, %xmm5
+; AVX1-NEXT:    vpmullw %xmm3, %xmm6, %xmm3
+; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm6
+; AVX1-NEXT:    vpackuswb %xmm6, %xmm4, %xmm4
+; AVX1-NEXT:    vpand %xmm5, %xmm1, %xmm1
+; AVX1-NEXT:    vpand %xmm5, %xmm3, %xmm3
+; AVX1-NEXT:    vpackuswb %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vpcmpgtb %xmm3, %xmm10, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovsxbw %xmm2, %xmm4
 ; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm6
-; AVX1-NEXT:    vpmullw %xmm5, %xmm6, %xmm5
+; AVX1-NEXT:    vpmullw %xmm4, %xmm6, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm6
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm0
 ; AVX1-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm5, %xmm2
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpcmpgtb %xmm7, %xmm11, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm6, %xmm2
+; AVX1-NEXT:    vpand %xmm5, %xmm4, %xmm4
+; AVX1-NEXT:    vpand %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm4, %xmm4
+; AVX1-NEXT:    vpcmpgtb %xmm4, %xmm10, %xmm0
 ; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
-; AVX1-NEXT:    vpxor %xmm5, %xmm8, %xmm6
-; AVX1-NEXT:    vpxor %xmm5, %xmm3, %xmm3
-; AVX1-NEXT:    vpxor %xmm5, %xmm4, %xmm2
-; AVX1-NEXT:    vpxor %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm7, %xmm7, %xmm7
+; AVX1-NEXT:    vpxor %xmm7, %xmm8, %xmm6
+; AVX1-NEXT:    vpxor %xmm7, %xmm12, %xmm5
+; AVX1-NEXT:    vpxor %xmm7, %xmm1, %xmm2
+; AVX1-NEXT:    vpxor %xmm7, %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqa %xmm9, 48(%rsi)
-; AVX1-NEXT:    vmovdqa %xmm12, 32(%rsi)
-; AVX1-NEXT:    vmovdqa %xmm1, 16(%rsi)
-; AVX1-NEXT:    vmovdqa %xmm7, (%rsi)
+; AVX1-NEXT:    vmovdqa %xmm11, 32(%rsi)
+; AVX1-NEXT:    vmovdqa %xmm3, 16(%rsi)
+; AVX1-NEXT:    vmovdqa %xmm4, (%rsi)
 ; AVX1-NEXT:    vpmovsxbd %xmm6, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 192(%rdi)
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm1
+; AVX1-NEXT:    vpmovsxbd %xmm5, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 128(%rdi)
 ; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 64(%rdi)
 ; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, (%rdi)
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 224(%rdi)
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 240(%rdi)
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 208(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 160(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm3[3,3,3,3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 176(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 144(%rdi)
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 96(%rdi)
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vmovdqa %xmm1, 112(%rdi)

 ; AVX1-NEXT:    vmovdqa %xmm0, 16(%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: smulo_v64i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    movq %rdi, %rax
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm4 = ymm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX2-NEXT:    vpsraw $8, %ymm4, %ymm4
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX2-NEXT:    vpsraw $8, %ymm5, %ymm5
 ; AVX2-NEXT:    vpmullw %ymm4, %ymm5, %ymm4
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpand %ymm5, %ymm4, %ymm4
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm6 = ymm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm7 = ymm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpmullw %ymm6, %ymm7, %ymm6
-; AVX2-NEXT:    vpand %ymm5, %ymm6, %ymm6
-; AVX2-NEXT:    vpackuswb %ymm4, %ymm6, %ymm10
-; AVX2-NEXT:    vpxor %xmm6, %xmm6, %xmm6
-; AVX2-NEXT:    vpcmpgtb %ymm10, %ymm6, %ymm7
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm8 = ymm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpsraw $8, %ymm8, %ymm8
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm9 = ymm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpsraw $8, %ymm9, %ymm9
-; AVX2-NEXT:    vpmullw %ymm8, %ymm9, %ymm8
-; AVX2-NEXT:    vpsrlw $8, %ymm8, %ymm8
+; AVX2-NEXT:    vpsrlw $8, %ymm4, %ymm5
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm3 = ymm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsraw $8, %ymm3, %ymm3
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsraw $8, %ymm1, %ymm1
 ; AVX2-NEXT:    vpmullw %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vpsrlw $8, %ymm1, %ymm1
-; AVX2-NEXT:    vpackuswb %ymm8, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqb %ymm7, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm7, %ymm7, %ymm7
-; AVX2-NEXT:    vpxor %ymm7, %ymm1, %ymm1
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm8 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpmullw %ymm3, %ymm8, %ymm3
-; AVX2-NEXT:    vpand %ymm5, %ymm3, %ymm3
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm8 = ymm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm9 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpmullw %ymm8, %ymm9, %ymm8
-; AVX2-NEXT:    vpand %ymm5, %ymm8, %ymm5
-; AVX2-NEXT:    vpackuswb %ymm3, %ymm5, %ymm3
-; AVX2-NEXT:    vpcmpgtb %ymm3, %ymm6, %ymm5
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm6 = ymm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpsraw $8, %ymm6, %ymm6
+; AVX2-NEXT:    vpsrlw $8, %ymm1, %ymm3
+; AVX2-NEXT:    vpackuswb %ymm5, %ymm3, %ymm3
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-NEXT:    vpand %ymm5, %ymm4, %ymm4
+; AVX2-NEXT:    vpand %ymm5, %ymm1, %ymm1
+; AVX2-NEXT:    vpackuswb %ymm4, %ymm1, %ymm9
+; AVX2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX2-NEXT:    vpcmpgtb %ymm9, %ymm4, %ymm6
+; AVX2-NEXT:    vpcmpeqb %ymm3, %ymm6, %ymm3
+; AVX2-NEXT:    vpcmpeqd %ymm6, %ymm6, %ymm6
+; AVX2-NEXT:    vpxor %ymm6, %ymm3, %ymm3
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm7 = ymm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
+; AVX2-NEXT:    vpsraw $8, %ymm7, %ymm7
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm8 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
 ; AVX2-NEXT:    vpsraw $8, %ymm8, %ymm8
-; AVX2-NEXT:    vpmullw %ymm6, %ymm8, %ymm6
-; AVX2-NEXT:    vpsrlw $8, %ymm6, %ymm6
+; AVX2-NEXT:    vpmullw %ymm7, %ymm8, %ymm7
+; AVX2-NEXT:    vpsrlw $8, %ymm7, %ymm8
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm2 = ymm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsraw $8, %ymm2, %ymm2
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
 ; AVX2-NEXT:    vpsraw $8, %ymm0, %ymm0
 ; AVX2-NEXT:    vpmullw %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
-; AVX2-NEXT:    vpackuswb %ymm6, %ymm0, %ymm0
-; AVX2-NEXT:    vpcmpeqb %ymm5, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %ymm7, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
-; AVX2-NEXT:    vpmovsxbd %xmm2, %ymm8
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm5
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm2
+; AVX2-NEXT:    vpackuswb %ymm8, %ymm2, %ymm2
+; AVX2-NEXT:    vpand %ymm5, %ymm7, %ymm7
+; AVX2-NEXT:    vpand %ymm5, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm7, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtb %ymm0, %ymm4, %ymm4
+; AVX2-NEXT:    vpcmpeqb %ymm2, %ymm4, %ymm2
+; AVX2-NEXT:    vpxor %ymm6, %ymm2, %ymm2
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[2,3,2,3]
+; AVX2-NEXT:    vpmovsxbd %xmm4, %ymm8
+; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm5
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm6 = xmm5[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm6, %ymm6
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm7 = xmm1[2,3,2,3]
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm7 = xmm3[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm7, %ymm7
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm4
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm4[2,3,2,3]
+; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm4
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm1
 ; AVX2-NEXT:    vpmovsxbd %xmm2, %ymm2
-; AVX2-NEXT:    vpmovsxbd %xmm0, %ymm0
 ; AVX2-NEXT:    vpmovsxbd %xmm5, %ymm5
-; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm1
+; AVX2-NEXT:    vpmovsxbd %xmm3, %ymm3
 ; AVX2-NEXT:    vpmovsxbd %xmm4, %ymm4
-; AVX2-NEXT:    vmovdqa %ymm10, 32(%rsi)
-; AVX2-NEXT:    vmovdqa %ymm3, (%rsi)
+; AVX2-NEXT:    vmovdqa %ymm9, 32(%rsi)
+; AVX2-NEXT:    vmovdqa %ymm0, (%rsi)
 ; AVX2-NEXT:    vmovdqa %ymm4, 192(%rdi)
-; AVX2-NEXT:    vmovdqa %ymm1, 128(%rdi)
+; AVX2-NEXT:    vmovdqa %ymm3, 128(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm5, 64(%rdi)
-; AVX2-NEXT:    vmovdqa %ymm0, (%rdi)
-; AVX2-NEXT:    vmovdqa %ymm2, 224(%rdi)
+; AVX2-NEXT:    vmovdqa %ymm2, (%rdi)
+; AVX2-NEXT:    vmovdqa %ymm1, 224(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm7, 160(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm6, 96(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm8, 32(%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: smulo_v64i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
-; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm5
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero,xmm5[8],zero,xmm5[9],zero,xmm5[10],zero,xmm5[11],zero,xmm5[12],zero,xmm5[13],zero,xmm5[14],zero,xmm5[15],zero
-; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm6
-; AVX512F-NEXT:    vextracti128 $1, %ymm6, %xmm7
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm4 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero,xmm7[8],zero,xmm7[9],zero,xmm7[10],zero,xmm7[11],zero,xmm7[12],zero,xmm7[13],zero,xmm7[14],zero,xmm7[15],zero
-; AVX512F-NEXT:    vpmullw %ymm2, %ymm4, %ymm2
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
+; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm2
+; AVX512F-NEXT:    vpmovsxbw %xmm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm5
+; AVX512F-NEXT:    vextracti128 $1, %ymm5, %xmm4
+; AVX512F-NEXT:    vpmovsxbw %xmm4, %ymm4
+; AVX512F-NEXT:    vpmullw %ymm2, %ymm4, %ymm6
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm6[0],zero,ymm6[1],zero,ymm6[2],zero,ymm6[3],zero,ymm6[4],zero,ymm6[5],zero,ymm6[6],zero,ymm6[7],zero,ymm6[8],zero,ymm6[9],zero,ymm6[10],zero,ymm6[11],zero,ymm6[12],zero,ymm6[13],zero,ymm6[14],zero,ymm6[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm2, %xmm4
-; AVX512F-NEXT:    vpxor %xmm8, %xmm8, %xmm8
-; AVX512F-NEXT:    vpcmpgtb %xmm4, %xmm8, %xmm2
+; AVX512F-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512F-NEXT:    vpcmpgtb %xmm4, %xmm2, %xmm7
+; AVX512F-NEXT:    vpsrlw $8, %ymm6, %ymm6
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm6 = ymm6[0],zero,ymm6[1],zero,ymm6[2],zero,ymm6[3],zero,ymm6[4],zero,ymm6[5],zero,ymm6[6],zero,ymm6[7],zero,ymm6[8],zero,ymm6[9],zero,ymm6[10],zero,ymm6[11],zero,ymm6[12],zero,ymm6[13],zero,ymm6[14],zero,ymm6[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm6, %xmm6
+; AVX512F-NEXT:    vpcmpeqb %xmm6, %xmm7, %xmm6
+; AVX512F-NEXT:    vpternlogq $15, %xmm6, %xmm6, %xmm6
+; AVX512F-NEXT:    vpmovsxbd %xmm6, %zmm6
+; AVX512F-NEXT:    vptestmd %zmm6, %zmm6, %k1
+; AVX512F-NEXT:    vpmovsxbw %xmm3, %ymm3
 ; AVX512F-NEXT:    vpmovsxbw %xmm5, %ymm5
-; AVX512F-NEXT:    vpmovsxbw %xmm7, %ymm7
-; AVX512F-NEXT:    vpmullw %ymm5, %ymm7, %ymm5
-; AVX512F-NEXT:    vpsrlw $8, %ymm5, %ymm5
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm5 = ymm5[0],zero,ymm5[1],zero,ymm5[2],zero,ymm5[3],zero,ymm5[4],zero,ymm5[5],zero,ymm5[6],zero,ymm5[7],zero,ymm5[8],zero,ymm5[9],zero,ymm5[10],zero,ymm5[11],zero,ymm5[12],zero,ymm5[13],zero,ymm5[14],zero,ymm5[15],zero
+; AVX512F-NEXT:    vpmullw %ymm3, %ymm5, %ymm3
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm5 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm5, %xmm5
-; AVX512F-NEXT:    vpcmpeqb %xmm2, %xmm5, %xmm2
-; AVX512F-NEXT:    vpternlogq $15, %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT:    vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k1
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero,xmm3[8],zero,xmm3[9],zero,xmm3[10],zero,xmm3[11],zero,xmm3[12],zero,xmm3[13],zero,xmm3[14],zero,xmm3[15],zero
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm5 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero,xmm6[4],zero,xmm6[5],zero,xmm6[6],zero,xmm6[7],zero,xmm6[8],zero,xmm6[9],zero,xmm6[10],zero,xmm6[11],zero,xmm6[12],zero,xmm6[13],zero,xmm6[14],zero,xmm6[15],zero
-; AVX512F-NEXT:    vpmullw %ymm2, %ymm5, %ymm2
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm2, %xmm5
-; AVX512F-NEXT:    vpcmpgtb %xmm5, %xmm8, %xmm2
+; AVX512F-NEXT:    vpcmpgtb %xmm5, %xmm2, %xmm6
+; AVX512F-NEXT:    vpsrlw $8, %ymm3, %ymm3
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm3, %xmm3
+; AVX512F-NEXT:    vpcmpeqb %xmm3, %xmm6, %xmm3
+; AVX512F-NEXT:    vpternlogq $15, %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpmovsxbd %xmm3, %zmm3
+; AVX512F-NEXT:    vptestmd %zmm3, %zmm3, %k2
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm3
 ; AVX512F-NEXT:    vpmovsxbw %xmm3, %ymm3
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm6
 ; AVX512F-NEXT:    vpmovsxbw %xmm6, %ymm6
 ; AVX512F-NEXT:    vpmullw %ymm3, %ymm6, %ymm3
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm6 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm6, %xmm6
+; AVX512F-NEXT:    vpcmpgtb %xmm6, %xmm2, %xmm7
 ; AVX512F-NEXT:    vpsrlw $8, %ymm3, %ymm3
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm3, %xmm3
-; AVX512F-NEXT:    vpcmpeqb %xmm2, %xmm3, %xmm2
-; AVX512F-NEXT:    vpternlogq $15, %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT:    vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k2
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
-; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm7
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm6 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero,xmm7[8],zero,xmm7[9],zero,xmm7[10],zero,xmm7[11],zero,xmm7[12],zero,xmm7[13],zero,xmm7[14],zero,xmm7[15],zero
-; AVX512F-NEXT:    vpmullw %ymm3, %ymm6, %ymm3
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm3 = ymm3[0],zero,ymm3[1],zero,ymm3[2],zero,ymm3[3],zero,ymm3[4],zero,ymm3[5],zero,ymm3[6],zero,ymm3[7],zero,ymm3[8],zero,ymm3[9],zero,ymm3[10],zero,ymm3[11],zero,ymm3[12],zero,ymm3[13],zero,ymm3[14],zero,ymm3[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm3, %xmm6
-; AVX512F-NEXT:    vpcmpgtb %xmm6, %xmm8, %xmm3
-; AVX512F-NEXT:    vpmovsxbw %xmm2, %ymm2
-; AVX512F-NEXT:    vpmovsxbw %xmm7, %ymm7
-; AVX512F-NEXT:    vpmullw %ymm2, %ymm7, %ymm2
-; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm2, %xmm2
-; AVX512F-NEXT:    vpcmpeqb %xmm3, %xmm2, %xmm2
-; AVX512F-NEXT:    vpternlogq $15, %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT:    vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k3
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
-; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX512F-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
-; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm2 = ymm2[0],zero,ymm2[1],zero,ymm2[2],zero,ymm2[3],zero,ymm2[4],zero,ymm2[5],zero,ymm2[6],zero,ymm2[7],zero,ymm2[8],zero,ymm2[9],zero,ymm2[10],zero,ymm2[11],zero,ymm2[12],zero,ymm2[13],zero,ymm2[14],zero,ymm2[15],zero
-; AVX512F-NEXT:    vpmovdb %zmm2, %xmm7
-; AVX512F-NEXT:    vpcmpgtb %xmm7, %xmm8, %xmm2
+; AVX512F-NEXT:    vpcmpeqb %xmm3, %xmm7, %xmm3
+; AVX512F-NEXT:    vpternlogq $15, %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpmovsxbd %xmm3, %zmm3
+; AVX512F-NEXT:    vptestmd %zmm3, %zmm3, %k3
 ; AVX512F-NEXT:    vpmovsxbw %xmm1, %ymm1
 ; AVX512F-NEXT:    vpmovsxbw %xmm0, %ymm0
 ; AVX512F-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512F-NEXT:    vpmovdb %zmm1, %xmm7
+; AVX512F-NEXT:    vpcmpgtb %xmm7, %xmm2, %xmm1
 ; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} 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,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT:    vpcmpeqb %xmm0, %xmm1, %xmm0
 ; AVX512F-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
 ; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k4
 ; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k4} {z}
 ; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k3} {z}
 ; AVX512F-NEXT:    vpternlogd $255, %zmm2, %zmm2, %zmm2 {%k2} {z}
 ; AVX512F-NEXT:    vpternlogd $255, %zmm3, %zmm3, %zmm3 {%k1} {z}
 ; AVX512F-NEXT:    vmovdqa %xmm4, 48(%rdi)
 ; AVX512F-NEXT:    vmovdqa %xmm5, 32(%rdi)
 ; AVX512F-NEXT:    vmovdqa %xmm6, 16(%rdi)
 ; AVX512F-NEXT:    vmovdqa %xmm7, (%rdi)
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512BW-LABEL: smulo_v64i8:
 ; AVX512BW:       # %bb.0:
 ; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm2 = zmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
 ; AVX512BW-NEXT:    vpsraw $8, %zmm2, %zmm2
 ; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm3 = zmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
 ; AVX512BW-NEXT:    vpsraw $8, %zmm3, %zmm3
 ; AVX512BW-NEXT:    vpmullw %zmm2, %zmm3, %zmm2
-; AVX512BW-NEXT:    vpsrlw $8, %zmm2, %zmm2
-; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm3 = zmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
-; AVX512BW-NEXT:    vpsraw $8, %zmm3, %zmm3
-; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm4 = zmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
-; AVX512BW-NEXT:    vpsraw $8, %zmm4, %zmm4
-; AVX512BW-NEXT:    vpmullw %zmm3, %zmm4, %zmm3
-; AVX512BW-NEXT:    vpsrlw $8, %zmm3, %zmm3
-; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm3, %zmm2
-; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm3 = zmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
-; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm4 = zmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
-; AVX512BW-NEXT:    vpmullw %zmm3, %zmm4, %zmm3
-; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512BW-NEXT:    vpandq %zmm4, %zmm3, %zmm3
+; AVX512BW-NEXT:    vpsrlw $8, %zmm2, %zmm3
 ; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm1 = zmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
+; AVX512BW-NEXT:    vpsraw $8, %zmm1, %zmm1
 ; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm0 = zmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
+; AVX512BW-NEXT:    vpsraw $8, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vpmullw %zmm1, %zmm0, %zmm0
-; AVX512BW-NEXT:    vpandq %zmm4, %zmm0, %zmm0
-; AVX512BW-NEXT:    vpackuswb %zmm3, %zmm0, %zmm4
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpackuswb %zmm3, %zmm1, %zmm1
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512BW-NEXT:    vpandq %zmm3, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpandq %zmm3, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm4
 ; AVX512BW-NEXT:    vpmovb2m %zmm4, %k0
 ; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
-; AVX512BW-NEXT:    vpcmpneqb %zmm0, %zmm2, %k1
+; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k2
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k2} {z}
 ; AVX512BW-NEXT:    kshiftrq $32, %k1, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm3, %zmm3, %zmm3 {%k1} {z}
 ; AVX512BW-NEXT:    vmovdqa64 %zmm4, (%rdi)

llvm/test/CodeGen/X86/vec_umulo.ll

; AVX512-NEXT:    retq
   %res = sext <16 x i1> %obit to <16 x i32>
   store <16 x i32> %val, <16 x i32>* %p2
   ret <16 x i32> %res
 }
 
 define <16 x i32> @umulo_v16i8(<16 x i8> %a0, <16 x i8> %a1, <16 x i8>* %p2) nounwind {
 ; SSE2-LABEL: umulo_v16i8:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa %xmm1, %xmm2
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm3
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm2, %xmm3
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    pand %xmm2, %xmm3
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm5, %xmm4
-; SSE2-NEXT:    pand %xmm2, %xmm4
-; SSE2-NEXT:    packuswb %xmm3, %xmm4
 ; SSE2-NEXT:    pxor %xmm2, %xmm2
 ; SSE2-NEXT:    movdqa %xmm1, %xmm3
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm5
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm2[8],xmm5[9],xmm2[9],xmm5[10],xmm2[10],xmm5[11],xmm2[11],xmm5[12],xmm2[12],xmm5[13],xmm2[13],xmm5[14],xmm2[14],xmm5[15],xmm2[15]
 ; SSE2-NEXT:    pmullw %xmm3, %xmm5
-; SSE2-NEXT:    psrlw $8, %xmm5
+; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
+; SSE2-NEXT:    movdqa %xmm5, %xmm3
+; SSE2-NEXT:    pand %xmm4, %xmm3
 ; 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:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; SSE2-NEXT:    pmullw %xmm1, %xmm0
+; SSE2-NEXT:    pand %xmm0, %xmm4
+; SSE2-NEXT:    packuswb %xmm3, %xmm4
+; SSE2-NEXT:    psrlw $8, %xmm5
 ; SSE2-NEXT:    psrlw $8, %xmm0
 ; SSE2-NEXT:    packuswb %xmm5, %xmm0
 ; SSE2-NEXT:    pcmpeqb %xmm2, %xmm0
 ; SSE2-NEXT:    pcmpeqd %xmm3, %xmm3
 ; SSE2-NEXT:    pxor %xmm0, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, %xmm1
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    movdqa %xmm1, %xmm0

 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
 ; SSE2-NEXT:    movdqa %xmm4, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: umulo_v16i8:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    movdqa %xmm1, %xmm2
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm3
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm2, %xmm3
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
-; SSSE3-NEXT:    pand %xmm2, %xmm3
-; SSSE3-NEXT:    movdqa %xmm1, %xmm5
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm4
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm5, %xmm4
-; SSSE3-NEXT:    pand %xmm2, %xmm4
-; SSSE3-NEXT:    packuswb %xmm3, %xmm4
 ; SSSE3-NEXT:    pxor %xmm2, %xmm2
 ; SSSE3-NEXT:    movdqa %xmm1, %xmm3
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
 ; SSSE3-NEXT:    movdqa %xmm0, %xmm5
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm2[8],xmm5[9],xmm2[9],xmm5[10],xmm2[10],xmm5[11],xmm2[11],xmm5[12],xmm2[12],xmm5[13],xmm2[13],xmm5[14],xmm2[14],xmm5[15],xmm2[15]
 ; SSSE3-NEXT:    pmullw %xmm3, %xmm5
-; SSSE3-NEXT:    psrlw $8, %xmm5
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
+; SSSE3-NEXT:    movdqa %xmm5, %xmm3
+; SSSE3-NEXT:    pand %xmm4, %xmm3
 ; SSSE3-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]
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
 ; SSSE3-NEXT:    pmullw %xmm1, %xmm0
+; SSSE3-NEXT:    pand %xmm0, %xmm4
+; SSSE3-NEXT:    packuswb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $8, %xmm5
 ; SSSE3-NEXT:    psrlw $8, %xmm0
 ; SSSE3-NEXT:    packuswb %xmm5, %xmm0
 ; SSSE3-NEXT:    pcmpeqb %xmm2, %xmm0
 ; SSSE3-NEXT:    pcmpeqd %xmm3, %xmm3
 ; SSSE3-NEXT:    pxor %xmm0, %xmm3
 ; SSSE3-NEXT:    movdqa %xmm3, %xmm1
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    movdqa %xmm1, %xmm0

 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
 ; SSSE3-NEXT:    movdqa %xmm4, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: umulo_v16i8:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa %xmm1, %xmm2
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm0, %xmm3
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm2, %xmm3
-; SSE41-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
-; SSE41-NEXT:    pand %xmm4, %xmm3
-; 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:    pmovzxbw {{.*#+}} 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
-; SSE41-NEXT:    pmullw %xmm2, %xmm5
-; SSE41-NEXT:    pand %xmm5, %xmm4
-; SSE41-NEXT:    packuswb %xmm3, %xmm4
 ; SSE41-NEXT:    pxor %xmm2, %xmm2
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm3 = 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],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]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} 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
 ; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; SSE41-NEXT:    pmullw %xmm1, %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
+; SSE41-NEXT:    movdqa %xmm0, %xmm1
+; SSE41-NEXT:    pand %xmm4, %xmm1
+; SSE41-NEXT:    pmullw %xmm3, %xmm5
+; SSE41-NEXT:    pand %xmm5, %xmm4
+; SSE41-NEXT:    packuswb %xmm1, %xmm4
 ; SSE41-NEXT:    psrlw $8, %xmm0
 ; SSE41-NEXT:    psrlw $8, %xmm5
 ; SSE41-NEXT:    packuswb %xmm0, %xmm5
 ; SSE41-NEXT:    pcmpeqb %xmm2, %xmm5
 ; SSE41-NEXT:    pcmpeqd %xmm3, %xmm3
 ; SSE41-NEXT:    pxor %xmm5, %xmm3
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0

 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm3
 ; SSE41-NEXT:    psrad $31, %xmm3
 ; SSE41-NEXT:    movdqa %xmm4, (%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: umulo_v16i8:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT:    vpand %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; 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:    vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpand %xmm3, %xmm4, %xmm3
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm3[8],xmm0[9],xmm3[9],xmm0[10],xmm3[10],xmm0[11],xmm3[11],xmm0[12],xmm3[12],xmm0[13],xmm3[13],xmm0[14],xmm3[14],xmm0[15],xmm3[15]
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = 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]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; AVX1-NEXT:    vpmullw %xmm3, %xmm4, %xmm3
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm4, %xmm3, %xmm5
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; 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:    vpand %xmm4, %xmm0, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm1, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm1
 ; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm1
-; AVX1-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
-; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
+; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vmovdqa %xmm2, (%rdi)
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT:    vmovdqa %xmm4, (%rdi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: umulo_v16i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
 ; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
 ; AVX2-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm1
; AVX512BW-NEXT:    retq
   store <16 x i8> %val, <16 x i8>* %p2
   ret <16 x i32> %res
 }
 
 define <32 x i32> @umulo_v32i8(<32 x i8> %a0, <32 x i8> %a1, <32 x i8>* %p2) nounwind {
 ; SSE2-LABEL: umulo_v32i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movq %rdi, %rax
+; SSE2-NEXT:    pxor %xmm5, %xmm5
 ; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm4, %xmm5
-; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    pand %xmm8, %xmm5
-; SSE2-NEXT:    movdqa %xmm2, %xmm7
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm0, %xmm9
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm7, %xmm9
-; SSE2-NEXT:    pand %xmm8, %xmm9
-; SSE2-NEXT:    packuswb %xmm5, %xmm9
-; SSE2-NEXT:    movdqa %xmm3, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm5, %xmm7
-; SSE2-NEXT:    pand %xmm8, %xmm7
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm1, %xmm10
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm6, %xmm10
-; SSE2-NEXT:    pand %xmm8, %xmm10
-; SSE2-NEXT:    packuswb %xmm7, %xmm10
-; SSE2-NEXT:    pxor %xmm6, %xmm6
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; SSE2-NEXT:    movdqa %xmm0, %xmm6
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
+; SSE2-NEXT:    pmullw %xmm4, %xmm6
+; SSE2-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
+; SSE2-NEXT:    movdqa %xmm6, %xmm7
+; SSE2-NEXT:    pand %xmm9, %xmm7
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE2-NEXT:    pmullw %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm8
+; SSE2-NEXT:    pand %xmm9, %xmm8
+; SSE2-NEXT:    packuswb %xmm7, %xmm8
 ; SSE2-NEXT:    movdqa %xmm3, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm4
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSE2-NEXT:    pmullw %xmm7, %xmm4
-; SSE2-NEXT:    psrlw $8, %xmm4
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSE2-NEXT:    pmullw %xmm7, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
+; SSE2-NEXT:    pand %xmm9, %xmm7
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
 ; SSE2-NEXT:    pmullw %xmm3, %xmm1
+; SSE2-NEXT:    pand %xmm1, %xmm9
+; SSE2-NEXT:    packuswb %xmm7, %xmm9
+; SSE2-NEXT:    psrlw $8, %xmm2
 ; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    packuswb %xmm4, %xmm1
-; SSE2-NEXT:    pcmpeqb %xmm6, %xmm1
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSE2-NEXT:    pxor %xmm3, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSE2-NEXT:    pmullw %xmm4, %xmm7
-; SSE2-NEXT:    psrlw $8, %xmm7
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE2-NEXT:    pmullw %xmm2, %xmm0
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpeqb %xmm5, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE2-NEXT:    pxor %xmm2, %xmm1
+; SSE2-NEXT:    psrlw $8, %xmm6
 ; SSE2-NEXT:    psrlw $8, %xmm0
-; SSE2-NEXT:    packuswb %xmm7, %xmm0
-; SSE2-NEXT:    pcmpeqb %xmm6, %xmm0
-; SSE2-NEXT:    pxor %xmm3, %xmm0
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    packuswb %xmm6, %xmm0
+; SSE2-NEXT:    pcmpeqb %xmm5, %xmm0
+; SSE2-NEXT:    pxor %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm3
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3]
+; SSE2-NEXT:    pslld $31, %xmm5
+; SSE2-NEXT:    psrad $31, %xmm5
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pslld $31, %xmm2
-; SSE2-NEXT:    psrad $31, %xmm2
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm6
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm6
 ; SSE2-NEXT:    psrad $31, %xmm6
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm0
 ; SSE2-NEXT:    psrad $31, %xmm0
-; SSE2-NEXT:    movdqa %xmm1, %xmm4
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm4, %xmm7
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    movdqa %xmm2, %xmm7
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm7
 ; SSE2-NEXT:    psrad $31, %xmm7
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pslld $31, %xmm2
+; SSE2-NEXT:    psrad $31, %xmm2
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm4
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm4
 ; SSE2-NEXT:    psrad $31, %xmm4
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm5
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3]
-; SSE2-NEXT:    pslld $31, %xmm5
-; SSE2-NEXT:    psrad $31, %xmm5
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm1
 ; SSE2-NEXT:    psrad $31, %xmm1
-; SSE2-NEXT:    movdqa %xmm10, 16(%rsi)
-; SSE2-NEXT:    movdqa %xmm9, (%rsi)
+; SSE2-NEXT:    movdqa %xmm9, 16(%rsi)
+; SSE2-NEXT:    movdqa %xmm8, (%rsi)
 ; SSE2-NEXT:    movdqa %xmm1, 112(%rdi)
-; SSE2-NEXT:    movdqa %xmm5, 96(%rdi)
-; SSE2-NEXT:    movdqa %xmm4, 80(%rdi)
+; SSE2-NEXT:    movdqa %xmm4, 96(%rdi)
+; SSE2-NEXT:    movdqa %xmm2, 80(%rdi)
 ; SSE2-NEXT:    movdqa %xmm7, 64(%rdi)
 ; SSE2-NEXT:    movdqa %xmm0, 48(%rdi)
 ; SSE2-NEXT:    movdqa %xmm6, 32(%rdi)
-; SSE2-NEXT:    movdqa %xmm2, 16(%rdi)
-; SSE2-NEXT:    movdqa %xmm3, (%rdi)
+; SSE2-NEXT:    movdqa %xmm3, 16(%rdi)
+; SSE2-NEXT:    movdqa %xmm5, (%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: umulo_v32i8:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    movq %rdi, %rax
+; SSSE3-NEXT:    pxor %xmm5, %xmm5
 ; SSSE3-NEXT:    movdqa %xmm2, %xmm4
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm5
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm4, %xmm5
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
-; SSSE3-NEXT:    pand %xmm8, %xmm5
-; SSSE3-NEXT:    movdqa %xmm2, %xmm7
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm9
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm7, %xmm9
-; SSSE3-NEXT:    pand %xmm8, %xmm9
-; SSSE3-NEXT:    packuswb %xmm5, %xmm9
-; SSSE3-NEXT:    movdqa %xmm3, %xmm5
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm5, %xmm7
-; SSSE3-NEXT:    pand %xmm8, %xmm7
-; SSSE3-NEXT:    movdqa %xmm3, %xmm6
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm10
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm6, %xmm10
-; SSSE3-NEXT:    pand %xmm8, %xmm10
-; SSSE3-NEXT:    packuswb %xmm7, %xmm10
-; SSSE3-NEXT:    pxor %xmm6, %xmm6
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm6
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
+; SSSE3-NEXT:    pmullw %xmm4, %xmm6
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
+; SSSE3-NEXT:    movdqa %xmm6, %xmm7
+; SSSE3-NEXT:    pand %xmm9, %xmm7
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSSE3-NEXT:    pmullw %xmm2, %xmm0
+; SSSE3-NEXT:    movdqa %xmm0, %xmm8
+; SSSE3-NEXT:    pand %xmm9, %xmm8
+; SSSE3-NEXT:    packuswb %xmm7, %xmm8
 ; SSSE3-NEXT:    movdqa %xmm3, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm4
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSSE3-NEXT:    pmullw %xmm7, %xmm4
-; SSSE3-NEXT:    psrlw $8, %xmm4
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSSE3-NEXT:    pmullw %xmm7, %xmm2
+; SSSE3-NEXT:    movdqa %xmm2, %xmm7
+; SSSE3-NEXT:    pand %xmm9, %xmm7
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
 ; SSSE3-NEXT:    pmullw %xmm3, %xmm1
+; SSSE3-NEXT:    pand %xmm1, %xmm9
+; SSSE3-NEXT:    packuswb %xmm7, %xmm9
+; SSSE3-NEXT:    psrlw $8, %xmm2
 ; SSSE3-NEXT:    psrlw $8, %xmm1
-; SSSE3-NEXT:    packuswb %xmm4, %xmm1
-; SSSE3-NEXT:    pcmpeqb %xmm6, %xmm1
-; SSSE3-NEXT:    pcmpeqd %xmm3, %xmm3
-; SSSE3-NEXT:    pxor %xmm3, %xmm1
-; SSSE3-NEXT:    movdqa %xmm2, %xmm4
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSSE3-NEXT:    pmullw %xmm4, %xmm7
-; SSSE3-NEXT:    psrlw $8, %xmm7
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSSE3-NEXT:    pmullw %xmm2, %xmm0
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpeqb %xmm5, %xmm1
+; SSSE3-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSSE3-NEXT:    pxor %xmm2, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm6
 ; SSSE3-NEXT:    psrlw $8, %xmm0
-; SSSE3-NEXT:    packuswb %xmm7, %xmm0
-; SSSE3-NEXT:    pcmpeqb %xmm6, %xmm0
-; SSSE3-NEXT:    pxor %xmm3, %xmm0
-; SSSE3-NEXT:    movdqa %xmm0, %xmm2
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm3
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    packuswb %xmm6, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm5, %xmm0
+; SSSE3-NEXT:    pxor %xmm2, %xmm0
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm5
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3]
+; SSSE3-NEXT:    pslld $31, %xmm5
+; SSSE3-NEXT:    psrad $31, %xmm5
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pslld $31, %xmm2
-; SSSE3-NEXT:    psrad $31, %xmm2
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; SSSE3-NEXT:    movdqa %xmm0, %xmm6
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm6
 ; SSSE3-NEXT:    psrad $31, %xmm6
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm0
 ; SSSE3-NEXT:    psrad $31, %xmm0
-; SSSE3-NEXT:    movdqa %xmm1, %xmm4
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm4, %xmm7
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    movdqa %xmm2, %xmm7
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm7
 ; SSSE3-NEXT:    psrad $31, %xmm7
-; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    pslld $31, %xmm2
+; SSSE3-NEXT:    psrad $31, %xmm2
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm4
 ; SSSE3-NEXT:    psrad $31, %xmm4
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm5
-; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3]
-; SSSE3-NEXT:    pslld $31, %xmm5
-; SSSE3-NEXT:    psrad $31, %xmm5
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm1
 ; SSSE3-NEXT:    psrad $31, %xmm1
-; SSSE3-NEXT:    movdqa %xmm10, 16(%rsi)
-; SSSE3-NEXT:    movdqa %xmm9, (%rsi)
+; SSSE3-NEXT:    movdqa %xmm9, 16(%rsi)
+; SSSE3-NEXT:    movdqa %xmm8, (%rsi)
 ; SSSE3-NEXT:    movdqa %xmm1, 112(%rdi)
-; SSSE3-NEXT:    movdqa %xmm5, 96(%rdi)
-; SSSE3-NEXT:    movdqa %xmm4, 80(%rdi)
+; SSSE3-NEXT:    movdqa %xmm4, 96(%rdi)
+; SSSE3-NEXT:    movdqa %xmm2, 80(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm7, 64(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm0, 48(%rdi)
 ; SSSE3-NEXT:    movdqa %xmm6, 32(%rdi)
-; SSSE3-NEXT:    movdqa %xmm2, 16(%rdi)
-; SSSE3-NEXT:    movdqa %xmm3, (%rdi)
+; SSSE3-NEXT:    movdqa %xmm3, 16(%rdi)
+; SSSE3-NEXT:    movdqa %xmm5, (%rdi)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: umulo_v32i8:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movq %rdi, %rax
-; SSE41-NEXT:    movdqa %xmm2, %xmm4
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm0, %xmm7
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm4, %xmm7
-; SSE41-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
-; SSE41-NEXT:    pand %xmm9, %xmm7
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm5 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; SSE41-NEXT:    pmovzxbw {{.*#+}} 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
-; SSE41-NEXT:    pmullw %xmm5, %xmm6
-; SSE41-NEXT:    movdqa %xmm6, %xmm8
-; SSE41-NEXT:    pand %xmm9, %xmm8
-; SSE41-NEXT:    packuswb %xmm7, %xmm8
-; SSE41-NEXT:    movdqa %xmm3, %xmm7
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm1, %xmm5
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm7, %xmm5
-; SSE41-NEXT:    pand %xmm9, %xmm5
-; 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:    pmovzxbw {{.*#+}} xmm7 = 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:    pmullw %xmm4, %xmm7
-; SSE41-NEXT:    pand %xmm7, %xmm9
-; SSE41-NEXT:    packuswb %xmm5, %xmm9
-; SSE41-NEXT:    pxor %xmm4, %xmm4
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
+; SSE41-NEXT:    pxor %xmm8, %xmm8
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; SSE41-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} 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
+; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE41-NEXT:    pmullw %xmm2, %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
+; SSE41-NEXT:    movdqa %xmm0, %xmm6
+; SSE41-NEXT:    pand %xmm2, %xmm6
+; SSE41-NEXT:    pmullw %xmm4, %xmm5
+; SSE41-NEXT:    movdqa %xmm5, %xmm9
+; SSE41-NEXT:    pand %xmm2, %xmm9
+; SSE41-NEXT:    packuswb %xmm6, %xmm9
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm7 = 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],xmm8[8],xmm3[9],xmm8[9],xmm3[10],xmm8[10],xmm3[11],xmm8[11],xmm3[12],xmm8[12],xmm3[13],xmm8[13],xmm3[14],xmm8[14],xmm3[15],xmm8[15]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm6 = 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],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
 ; SSE41-NEXT:    pmullw %xmm3, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, %xmm3
+; SSE41-NEXT:    pand %xmm2, %xmm3
+; SSE41-NEXT:    pmullw %xmm7, %xmm6
+; SSE41-NEXT:    pand %xmm6, %xmm2
+; SSE41-NEXT:    packuswb %xmm3, %xmm2
 ; SSE41-NEXT:    psrlw $8, %xmm1
-; SSE41-NEXT:    psrlw $8, %xmm7
-; SSE41-NEXT:    packuswb %xmm1, %xmm7
-; SSE41-NEXT:    pcmpeqb %xmm4, %xmm7
-; SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE41-NEXT:    pxor %xmm1, %xmm7
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
-; SSE41-NEXT:    pmullw %xmm2, %xmm0
-; SSE41-NEXT:    psrlw $8, %xmm0
 ; SSE41-NEXT:    psrlw $8, %xmm6
-; SSE41-NEXT:    packuswb %xmm0, %xmm6
-; SSE41-NEXT:    pcmpeqb %xmm4, %xmm6
+; SSE41-NEXT:    packuswb %xmm1, %xmm6
+; SSE41-NEXT:    pcmpeqb %xmm8, %xmm6
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
 ; SSE41-NEXT:    pxor %xmm1, %xmm6
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm0
-; SSE41-NEXT:    psrad $31, %xmm0
-; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm5
+; SSE41-NEXT:    packuswb %xmm0, %xmm5
+; SSE41-NEXT:    pcmpeqb %xmm8, %xmm5
+; SSE41-NEXT:    pxor %xmm1, %xmm5
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm8 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm8
+; SSE41-NEXT:    psrad $31, %xmm8
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero,xmm1[2],zero,zero,zero,xmm1[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm1
 ; SSE41-NEXT:    psrad $31, %xmm1
-; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[3,3,3,3]
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm2
-; SSE41-NEXT:    psrad $31, %xmm2
-; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[1,1,1,1]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm3 = xmm3[0],zero,zero,zero,xmm3[1],zero,zero,zero,xmm3[2],zero,zero,zero,xmm3[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm3
 ; SSE41-NEXT:    psrad $31, %xmm3
-; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[1,1,1,1]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm7 = xmm7[0],zero,zero,zero,xmm7[1],zero,zero,zero,xmm7[2],zero,zero,zero,xmm7[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm7
+; SSE41-NEXT:    psrad $31, %xmm7
+; SSE41-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm4 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero,xmm4[2],zero,zero,zero,xmm4[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm4
 ; SSE41-NEXT:    psrad $31, %xmm4
-; SSE41-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[3,3,3,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
+; SSE41-NEXT:    pslld $31, %xmm0
+; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm5 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm5
 ; SSE41-NEXT:    psrad $31, %xmm5
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm6 = xmm6[0],zero,zero,zero,xmm6[1],zero,zero,zero,xmm6[2],zero,zero,zero,xmm6[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm6
 ; SSE41-NEXT:    psrad $31, %xmm6
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm7 = xmm7[0],zero,zero,zero,xmm7[1],zero,zero,zero,xmm7[2],zero,zero,zero,xmm7[3],zero,zero,zero
-; SSE41-NEXT:    pslld $31, %xmm7
-; SSE41-NEXT:    psrad $31, %xmm7
-; SSE41-NEXT:    movdqa %xmm9, 16(%rsi)
-; SSE41-NEXT:    movdqa %xmm8, (%rsi)
-; SSE41-NEXT:    movdqa %xmm7, 64(%rdi)
-; SSE41-NEXT:    movdqa %xmm6, (%rdi)
-; SSE41-NEXT:    movdqa %xmm5, 112(%rdi)
+; SSE41-NEXT:    movdqa %xmm2, 16(%rsi)
+; SSE41-NEXT:    movdqa %xmm9, (%rsi)
+; SSE41-NEXT:    movdqa %xmm6, 64(%rdi)
+; SSE41-NEXT:    movdqa %xmm5, (%rdi)
+; SSE41-NEXT:    movdqa %xmm0, 112(%rdi)
 ; SSE41-NEXT:    movdqa %xmm4, 96(%rdi)
-; SSE41-NEXT:    movdqa %xmm3, 80(%rdi)
-; SSE41-NEXT:    movdqa %xmm2, 48(%rdi)
+; SSE41-NEXT:    movdqa %xmm7, 80(%rdi)
+; SSE41-NEXT:    movdqa %xmm3, 48(%rdi)
 ; SSE41-NEXT:    movdqa %xmm1, 32(%rdi)
-; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
+; SSE41-NEXT:    movdqa %xmm8, 16(%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: umulo_v32i8:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT:    vpand %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; 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:    vpmullw %xmm4, %xmm5, %xmm9
-; AVX1-NEXT:    vpand %xmm3, %xmm9, %xmm4
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm4, %xmm8
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm7
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm7[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm5, %xmm4, %xmm4
-; AVX1-NEXT:    vpand %xmm3, %xmm4, %xmm4
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm5 = 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:    vpmovzxbw {{.*#+}} xmm6 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero
-; AVX1-NEXT:    vpmullw %xmm5, %xmm6, %xmm6
-; AVX1-NEXT:    vpand %xmm3, %xmm6, %xmm3
-; AVX1-NEXT:    vpackuswb %xmm4, %xmm3, %xmm5
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm7[8],xmm3[8],xmm7[9],xmm3[9],xmm7[10],xmm3[10],xmm7[11],xmm3[11],xmm7[12],xmm3[12],xmm7[13],xmm3[13],xmm7[14],xmm3[14],xmm7[15],xmm3[15]
-; AVX1-NEXT:    vpmullw %xmm2, %xmm4, %xmm2
-; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrlw $8, %xmm6, %xmm4
-; AVX1-NEXT:    vpackuswb %xmm2, %xmm4, %xmm2
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT:    vpxor %xmm4, %xmm2, %xmm6
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm3[8],xmm0[9],xmm3[9],xmm0[10],xmm3[10],xmm0[11],xmm3[11],xmm0[12],xmm3[12],xmm0[13],xmm3[13],xmm0[14],xmm3[14],xmm0[15],xmm3[15]
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = 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]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; AVX1-NEXT:    vpmullw %xmm3, %xmm4, %xmm3
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm5, %xmm3, %xmm4
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm7 = 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 %xmm6, %xmm7, %xmm6
+; AVX1-NEXT:    vpand %xmm5, %xmm6, %xmm7
+; AVX1-NEXT:    vpackuswb %xmm4, %xmm7, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm7 = 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]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; AVX1-NEXT:    vpmullw %xmm7, %xmm4, %xmm4
+; AVX1-NEXT:    vpand %xmm5, %xmm4, %xmm7
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; 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:    vpand %xmm5, %xmm0, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm7, %xmm1, %xmm5
+; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm1
 ; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrlw $8, %xmm9, %xmm1
-; AVX1-NEXT:    vpackuswb %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm4, %xmm0, %xmm1
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm6, %xmm3
+; AVX1-NEXT:    vpackuswb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
 ; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-NEXT:    vpmovsxbd %xmm6, %xmm2
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm6[1,1,1,1]
+; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm4[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
 ; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm6[2,3,2,3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm6[3,3,3,3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[3,3,3,3]
 ; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm4
 ; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
 ; AVX1-NEXT:    vmovdqa %xmm5, 16(%rdi)
 ; AVX1-NEXT:    vmovdqa %xmm8, (%rdi)
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: umulo_v32i8:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm2 = ymm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpmullw %ymm2, %ymm3, %ymm2
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpand %ymm3, %ymm2, %ymm2
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm4 = ymm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm5 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpmullw %ymm4, %ymm5, %ymm4
-; AVX2-NEXT:    vpand %ymm3, %ymm4, %ymm3
-; AVX2-NEXT:    vpackuswb %ymm2, %ymm3, %ymm4
 ; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm3 = ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15],ymm1[24],ymm2[24],ymm1[25],ymm2[25],ymm1[26],ymm2[26],ymm1[27],ymm2[27],ymm1[28],ymm2[28],ymm1[29],ymm2[29],ymm1[30],ymm2[30],ymm1[31],ymm2[31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm0[8],ymm2[8],ymm0[9],ymm2[9],ymm0[10],ymm2[10],ymm0[11],ymm2[11],ymm0[12],ymm2[12],ymm0[13],ymm2[13],ymm0[14],ymm2[14],ymm0[15],ymm2[15],ymm0[24],ymm2[24],ymm0[25],ymm2[25],ymm0[26],ymm2[26],ymm0[27],ymm2[27],ymm0[28],ymm2[28],ymm0[29],ymm2[29],ymm0[30],ymm2[30],ymm0[31],ymm2[31]
-; AVX2-NEXT:    vpmullw %ymm3, %ymm5, %ymm3
-; AVX2-NEXT:    vpsrlw $8, %ymm3, %ymm3
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm4 = ymm0[8],ymm2[8],ymm0[9],ymm2[9],ymm0[10],ymm2[10],ymm0[11],ymm2[11],ymm0[12],ymm2[12],ymm0[13],ymm2[13],ymm0[14],ymm2[14],ymm0[15],ymm2[15],ymm0[24],ymm2[24],ymm0[25],ymm2[25],ymm0[26],ymm2[26],ymm0[27],ymm2[27],ymm0[28],ymm2[28],ymm0[29],ymm2[29],ymm0[30],ymm2[30],ymm0[31],ymm2[31]
+; AVX2-NEXT:    vpmullw %ymm3, %ymm4, %ymm3
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-NEXT:    vpand %ymm4, %ymm3, %ymm5
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[16],ymm2[16],ymm1[17],ymm2[17],ymm1[18],ymm2[18],ymm1[19],ymm2[19],ymm1[20],ymm2[20],ymm1[21],ymm2[21],ymm1[22],ymm2[22],ymm1[23],ymm2[23]
 ; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[2],ymm2[2],ymm0[3],ymm2[3],ymm0[4],ymm2[4],ymm0[5],ymm2[5],ymm0[6],ymm2[6],ymm0[7],ymm2[7],ymm0[16],ymm2[16],ymm0[17],ymm2[17],ymm0[18],ymm2[18],ymm0[19],ymm2[19],ymm0[20],ymm2[20],ymm0[21],ymm2[21],ymm0[22],ymm2[22],ymm0[23],ymm2[23]
 ; AVX2-NEXT:    vpmullw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm4, %ymm0, %ymm1
+; AVX2-NEXT:    vpackuswb %ymm5, %ymm1, %ymm4
+; AVX2-NEXT:    vpsrlw $8, %ymm3, %ymm1
 ; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
-; AVX2-NEXT:    vpackuswb %ymm3, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqb %ymm2, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
 ; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm1
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
 ; AVX2-NEXT:    vpmovsxbd %xmm3, %ymm2
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm1
; AVX512BW-NEXT:    retq
   store <32 x i8> %val, <32 x i8>* %p2
   ret <32 x i32> %res
 }
 
 define <64 x i32> @umulo_v64i8(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %p2) nounwind {
 ; SSE2-LABEL: umulo_v64i8:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movq %rdi, %rax
+; SSE2-NEXT:    pxor %xmm9, %xmm9
 ; SSE2-NEXT:    movdqa %xmm4, %xmm8
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm9
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm8, %xmm9
-; SSE2-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    pand %xmm12, %xmm9
-; SSE2-NEXT:    movdqa %xmm4, %xmm10
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm0, %xmm8
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm10, %xmm8
-; SSE2-NEXT:    pand %xmm12, %xmm8
-; SSE2-NEXT:    packuswb %xmm9, %xmm8
-; SSE2-NEXT:    movdqa %xmm5, %xmm9
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm9, %xmm10
-; SSE2-NEXT:    pand %xmm12, %xmm10
-; SSE2-NEXT:    movdqa %xmm5, %xmm11
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm1, %xmm9
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm11, %xmm9
-; SSE2-NEXT:    pand %xmm12, %xmm9
-; SSE2-NEXT:    packuswb %xmm10, %xmm9
-; SSE2-NEXT:    movdqa %xmm6, %xmm10
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm2, %xmm11
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm10, %xmm11
-; SSE2-NEXT:    pand %xmm12, %xmm11
-; SSE2-NEXT:    movdqa %xmm6, %xmm13
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm2, %xmm10
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm13, %xmm10
-; SSE2-NEXT:    pand %xmm12, %xmm10
-; SSE2-NEXT:    packuswb %xmm11, %xmm10
-; SSE2-NEXT:    movdqa %xmm7, %xmm11
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm13
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    pmullw %xmm11, %xmm13
-; SSE2-NEXT:    pand %xmm12, %xmm13
-; SSE2-NEXT:    movdqa %xmm7, %xmm14
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa %xmm3, %xmm11
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    pmullw %xmm14, %xmm11
-; SSE2-NEXT:    pand %xmm12, %xmm11
-; SSE2-NEXT:    packuswb %xmm13, %xmm11
-; SSE2-NEXT:    pxor %xmm12, %xmm12
-; SSE2-NEXT:    movdqa %xmm7, %xmm13
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm12[8],xmm13[9],xmm12[9],xmm13[10],xmm12[10],xmm13[11],xmm12[11],xmm13[12],xmm12[12],xmm13[13],xmm12[13],xmm13[14],xmm12[14],xmm13[15],xmm12[15]
-; SSE2-NEXT:    movdqa %xmm3, %xmm14
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm12[8],xmm14[9],xmm12[9],xmm14[10],xmm12[10],xmm14[11],xmm12[11],xmm14[12],xmm12[12],xmm14[13],xmm12[13],xmm14[14],xmm12[14],xmm14[15],xmm12[15]
-; SSE2-NEXT:    pmullw %xmm13, %xmm14
-; SSE2-NEXT:    psrlw $8, %xmm14
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1],xmm7[2],xmm12[2],xmm7[3],xmm12[3],xmm7[4],xmm12[4],xmm7[5],xmm12[5],xmm7[6],xmm12[6],xmm7[7],xmm12[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3],xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm9[8],xmm8[9],xmm9[9],xmm8[10],xmm9[10],xmm8[11],xmm9[11],xmm8[12],xmm9[12],xmm8[13],xmm9[13],xmm8[14],xmm9[14],xmm8[15],xmm9[15]
+; SSE2-NEXT:    movdqa %xmm0, %xmm10
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
+; SSE2-NEXT:    pmullw %xmm8, %xmm10
+; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
+; SSE2-NEXT:    movdqa %xmm10, %xmm12
+; SSE2-NEXT:    pand %xmm8, %xmm12
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE2-NEXT:    pmullw %xmm4, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm11
+; SSE2-NEXT:    pand %xmm8, %xmm11
+; SSE2-NEXT:    packuswb %xmm12, %xmm11
+; SSE2-NEXT:    movdqa %xmm5, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSE2-NEXT:    movdqa %xmm1, %xmm13
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm9[8],xmm13[9],xmm9[9],xmm13[10],xmm9[10],xmm13[11],xmm9[11],xmm13[12],xmm9[12],xmm13[13],xmm9[13],xmm13[14],xmm9[14],xmm13[15],xmm9[15]
+; SSE2-NEXT:    pmullw %xmm4, %xmm13
+; SSE2-NEXT:    movdqa %xmm13, %xmm4
+; SSE2-NEXT:    pand %xmm8, %xmm4
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSE2-NEXT:    pmullw %xmm5, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm12
+; SSE2-NEXT:    pand %xmm8, %xmm12
+; SSE2-NEXT:    packuswb %xmm4, %xmm12
+; SSE2-NEXT:    movdqa %xmm6, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm9[8],xmm5[9],xmm9[9],xmm5[10],xmm9[10],xmm5[11],xmm9[11],xmm5[12],xmm9[12],xmm5[13],xmm9[13],xmm5[14],xmm9[14],xmm5[15],xmm9[15]
+; SSE2-NEXT:    pmullw %xmm4, %xmm5
+; SSE2-NEXT:    movdqa %xmm5, %xmm4
+; SSE2-NEXT:    pand %xmm8, %xmm4
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3],xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
+; SSE2-NEXT:    pmullw %xmm6, %xmm2
+; SSE2-NEXT:    movdqa %xmm2, %xmm14
+; SSE2-NEXT:    pand %xmm8, %xmm14
+; SSE2-NEXT:    packuswb %xmm4, %xmm14
+; SSE2-NEXT:    movdqa %xmm7, %xmm4
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSE2-NEXT:    movdqa %xmm3, %xmm6
+; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm9[8],xmm6[9],xmm9[9],xmm6[10],xmm9[10],xmm6[11],xmm9[11],xmm6[12],xmm9[12],xmm6[13],xmm9[13],xmm6[14],xmm9[14],xmm6[15],xmm9[15]
+; SSE2-NEXT:    pmullw %xmm4, %xmm6
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
 ; SSE2-NEXT:    pmullw %xmm7, %xmm3
+; SSE2-NEXT:    movdqa %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm8, %xmm4
+; SSE2-NEXT:    pand %xmm3, %xmm8
+; SSE2-NEXT:    packuswb %xmm4, %xmm8
+; SSE2-NEXT:    psrlw $8, %xmm6
 ; SSE2-NEXT:    psrlw $8, %xmm3
-; SSE2-NEXT:    packuswb %xmm14, %xmm3
-; SSE2-NEXT:    movdqa %xmm6, %xmm13
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm12[8],xmm13[9],xmm12[9],xmm13[10],xmm12[10],xmm13[11],xmm12[11],xmm13[12],xmm12[12],xmm13[13],xmm12[13],xmm13[14],xmm12[14],xmm13[15],xmm12[15]
-; SSE2-NEXT:    movdqa %xmm2, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
-; SSE2-NEXT:    pmullw %xmm13, %xmm7
-; SSE2-NEXT:    psrlw $8, %xmm7
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3],xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
-; SSE2-NEXT:    pmullw %xmm6, %xmm2
+; SSE2-NEXT:    packuswb %xmm6, %xmm3
+; SSE2-NEXT:    psrlw $8, %xmm5
 ; SSE2-NEXT:    psrlw $8, %xmm2
-; SSE2-NEXT:    packuswb %xmm7, %xmm2
-; SSE2-NEXT:    movdqa %xmm5, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
-; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
-; SSE2-NEXT:    pmullw %xmm6, %xmm7
-; SSE2-NEXT:    psrlw $8, %xmm7
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3],xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
-; SSE2-NEXT:    pmullw %xmm5, %xmm1
+; SSE2-NEXT:    packuswb %xmm5, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm13
 ; SSE2-NEXT:    psrlw $8, %xmm1
-; SSE2-NEXT:    packuswb %xmm7, %xmm1
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm12[8],xmm5[9],xmm12[9],xmm5[10],xmm12[10],xmm5[11],xmm12[11],xmm5[12],xmm12[12],xmm5[13],xmm12[13],xmm5[14],xmm12[14],xmm5[15],xmm12[15]
-; SSE2-NEXT:    movdqa %xmm0, %xmm6
-; SSE2-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
-; SSE2-NEXT:    pmullw %xmm5, %xmm6
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3],xmm4[4],xmm12[4],xmm4[5],xmm12[5],xmm4[6],xmm12[6],xmm4[7],xmm12[7]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSE2-NEXT:    pmullw %xmm4, %xmm0
-; SSE2-NEXT:    psrlw $8, %xmm6
+; SSE2-NEXT:    packuswb %xmm13, %xmm1
+; SSE2-NEXT:    psrlw $8, %xmm10
 ; SSE2-NEXT:    psrlw $8, %xmm0
-; SSE2-NEXT:    packuswb %xmm6, %xmm0
-; SSE2-NEXT:    pcmpeqb %xmm12, %xmm3
-; SSE2-NEXT:    pcmpeqb %xmm12, %xmm2
-; SSE2-NEXT:    pcmpeqb %xmm12, %xmm1
-; SSE2-NEXT:    pcmpeqb %xmm12, %xmm0
+; SSE2-NEXT:    packuswb %xmm10, %xmm0
+; SSE2-NEXT:    pcmpeqb %xmm9, %xmm3
+; SSE2-NEXT:    pcmpeqb %xmm9, %xmm2
+; SSE2-NEXT:    pcmpeqb %xmm9, %xmm1
+; SSE2-NEXT:    pcmpeqb %xmm9, %xmm0
 ; SSE2-NEXT:    pcmpeqd %xmm4, %xmm4
 ; SSE2-NEXT:    pxor %xmm4, %xmm3
 ; SSE2-NEXT:    pxor %xmm4, %xmm2
 ; SSE2-NEXT:    pxor %xmm4, %xmm1
 ; SSE2-NEXT:    pxor %xmm4, %xmm0
-; SSE2-NEXT:    movdqa %xmm11, 48(%rsi)
-; SSE2-NEXT:    movdqa %xmm10, 32(%rsi)
-; SSE2-NEXT:    movdqa %xmm9, 16(%rsi)
+; SSE2-NEXT:    movdqa %xmm8, 48(%rsi)
+; SSE2-NEXT:    movdqa %xmm14, 32(%rsi)
+; SSE2-NEXT:    movdqa %xmm12, 16(%rsi)
 ; SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; SSE2-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE2-NEXT:    movdqa %xmm8, (%rsi)
+; SSE2-NEXT:    movdqa %xmm11, (%rsi)
 ; SSE2-NEXT:    movdqa %xmm3, %xmm5
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    pslld $31, %xmm3
 ; SSE2-NEXT:    psrad $31, %xmm3
 ; SSE2-NEXT:    movdqa %xmm3, 224(%rdi)
 ; SSE2-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
 ; SSE2-NEXT:    pslld $31, %xmm5

 ; SSE2-NEXT:    pslld $31, %xmm0
 ; SSE2-NEXT:    psrad $31, %xmm0
 ; SSE2-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: umulo_v64i8:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    movq %rdi, %rax
+; SSSE3-NEXT:    pxor %xmm9, %xmm9
 ; SSSE3-NEXT:    movdqa %xmm4, %xmm8
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm9
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm8, %xmm9
-; SSSE3-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,255,255,255]
-; SSSE3-NEXT:    pand %xmm12, %xmm9
-; SSSE3-NEXT:    movdqa %xmm4, %xmm10
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm8
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm10, %xmm8
-; SSSE3-NEXT:    pand %xmm12, %xmm8
-; SSSE3-NEXT:    packuswb %xmm9, %xmm8
-; SSSE3-NEXT:    movdqa %xmm5, %xmm9
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm10
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm9, %xmm10
-; SSSE3-NEXT:    pand %xmm12, %xmm10
-; SSSE3-NEXT:    movdqa %xmm5, %xmm11
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm9
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm11, %xmm9
-; SSSE3-NEXT:    pand %xmm12, %xmm9
-; SSSE3-NEXT:    packuswb %xmm10, %xmm9
-; SSSE3-NEXT:    movdqa %xmm6, %xmm10
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm10, %xmm11
-; SSSE3-NEXT:    pand %xmm12, %xmm11
-; SSSE3-NEXT:    movdqa %xmm6, %xmm13
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm10
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm13, %xmm10
-; SSSE3-NEXT:    pand %xmm12, %xmm10
-; SSSE3-NEXT:    packuswb %xmm11, %xmm10
-; SSSE3-NEXT:    movdqa %xmm7, %xmm11
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm3, %xmm13
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    pmullw %xmm11, %xmm13
-; SSSE3-NEXT:    pand %xmm12, %xmm13
-; SSSE3-NEXT:    movdqa %xmm7, %xmm14
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    movdqa %xmm3, %xmm11
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSSE3-NEXT:    pmullw %xmm14, %xmm11
-; SSSE3-NEXT:    pand %xmm12, %xmm11
-; SSSE3-NEXT:    packuswb %xmm13, %xmm11
-; SSSE3-NEXT:    pxor %xmm12, %xmm12
-; SSSE3-NEXT:    movdqa %xmm7, %xmm13
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm12[8],xmm13[9],xmm12[9],xmm13[10],xmm12[10],xmm13[11],xmm12[11],xmm13[12],xmm12[12],xmm13[13],xmm12[13],xmm13[14],xmm12[14],xmm13[15],xmm12[15]
-; SSSE3-NEXT:    movdqa %xmm3, %xmm14
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm12[8],xmm14[9],xmm12[9],xmm14[10],xmm12[10],xmm14[11],xmm12[11],xmm14[12],xmm12[12],xmm14[13],xmm12[13],xmm14[14],xmm12[14],xmm14[15],xmm12[15]
-; SSSE3-NEXT:    pmullw %xmm13, %xmm14
-; SSSE3-NEXT:    psrlw $8, %xmm14
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1],xmm7[2],xmm12[2],xmm7[3],xmm12[3],xmm7[4],xmm12[4],xmm7[5],xmm12[5],xmm7[6],xmm12[6],xmm7[7],xmm12[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3],xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm9[8],xmm8[9],xmm9[9],xmm8[10],xmm9[10],xmm8[11],xmm9[11],xmm8[12],xmm9[12],xmm8[13],xmm9[13],xmm8[14],xmm9[14],xmm8[15],xmm9[15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm10
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
+; SSSE3-NEXT:    pmullw %xmm8, %xmm10
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
+; SSSE3-NEXT:    movdqa %xmm10, %xmm12
+; SSSE3-NEXT:    pand %xmm8, %xmm12
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSSE3-NEXT:    pmullw %xmm4, %xmm0
+; SSSE3-NEXT:    movdqa %xmm0, %xmm11
+; SSSE3-NEXT:    pand %xmm8, %xmm11
+; SSSE3-NEXT:    packuswb %xmm12, %xmm11
+; SSSE3-NEXT:    movdqa %xmm5, %xmm4
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm13
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm9[8],xmm13[9],xmm9[9],xmm13[10],xmm9[10],xmm13[11],xmm9[11],xmm13[12],xmm9[12],xmm13[13],xmm9[13],xmm13[14],xmm9[14],xmm13[15],xmm9[15]
+; SSSE3-NEXT:    pmullw %xmm4, %xmm13
+; SSSE3-NEXT:    movdqa %xmm13, %xmm4
+; SSSE3-NEXT:    pand %xmm8, %xmm4
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSSE3-NEXT:    pmullw %xmm5, %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm12
+; SSSE3-NEXT:    pand %xmm8, %xmm12
+; SSSE3-NEXT:    packuswb %xmm4, %xmm12
+; SSSE3-NEXT:    movdqa %xmm6, %xmm4
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSSE3-NEXT:    movdqa %xmm2, %xmm5
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm9[8],xmm5[9],xmm9[9],xmm5[10],xmm9[10],xmm5[11],xmm9[11],xmm5[12],xmm9[12],xmm5[13],xmm9[13],xmm5[14],xmm9[14],xmm5[15],xmm9[15]
+; SSSE3-NEXT:    pmullw %xmm4, %xmm5
+; SSSE3-NEXT:    movdqa %xmm5, %xmm4
+; SSSE3-NEXT:    pand %xmm8, %xmm4
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3],xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
+; SSSE3-NEXT:    pmullw %xmm6, %xmm2
+; SSSE3-NEXT:    movdqa %xmm2, %xmm14
+; SSSE3-NEXT:    pand %xmm8, %xmm14
+; SSSE3-NEXT:    packuswb %xmm4, %xmm14
+; SSSE3-NEXT:    movdqa %xmm7, %xmm4
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm6
+; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm9[8],xmm6[9],xmm9[9],xmm6[10],xmm9[10],xmm6[11],xmm9[11],xmm6[12],xmm9[12],xmm6[13],xmm9[13],xmm6[14],xmm9[14],xmm6[15],xmm9[15]
+; SSSE3-NEXT:    pmullw %xmm4, %xmm6
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
+; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
 ; SSSE3-NEXT:    pmullw %xmm7, %xmm3
+; SSSE3-NEXT:    movdqa %xmm6, %xmm4
+; SSSE3-NEXT:    pand %xmm8, %xmm4
+; SSSE3-NEXT:    pand %xmm3, %xmm8
+; SSSE3-NEXT:    packuswb %xmm4, %xmm8
+; SSSE3-NEXT:    psrlw $8, %xmm6
 ; SSSE3-NEXT:    psrlw $8, %xmm3
-; SSSE3-NEXT:    packuswb %xmm14, %xmm3
-; SSSE3-NEXT:    movdqa %xmm6, %xmm13
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm12[8],xmm13[9],xmm12[9],xmm13[10],xmm12[10],xmm13[11],xmm12[11],xmm13[12],xmm12[12],xmm13[13],xmm12[13],xmm13[14],xmm12[14],xmm13[15],xmm12[15]
-; SSSE3-NEXT:    movdqa %xmm2, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
-; SSSE3-NEXT:    pmullw %xmm13, %xmm7
-; SSSE3-NEXT:    psrlw $8, %xmm7
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3],xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
-; SSSE3-NEXT:    pmullw %xmm6, %xmm2
+; SSSE3-NEXT:    packuswb %xmm6, %xmm3
+; SSSE3-NEXT:    psrlw $8, %xmm5
 ; SSSE3-NEXT:    psrlw $8, %xmm2
-; SSSE3-NEXT:    packuswb %xmm7, %xmm2
-; SSSE3-NEXT:    movdqa %xmm5, %xmm6
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm7
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
-; SSSE3-NEXT:    pmullw %xmm6, %xmm7
-; SSSE3-NEXT:    psrlw $8, %xmm7
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3],xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3],xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
-; SSSE3-NEXT:    pmullw %xmm5, %xmm1
+; SSSE3-NEXT:    packuswb %xmm5, %xmm2
+; SSSE3-NEXT:    psrlw $8, %xmm13
 ; SSSE3-NEXT:    psrlw $8, %xmm1
-; SSSE3-NEXT:    packuswb %xmm7, %xmm1
-; SSSE3-NEXT:    movdqa %xmm4, %xmm5
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm12[8],xmm5[9],xmm12[9],xmm5[10],xmm12[10],xmm5[11],xmm12[11],xmm5[12],xmm12[12],xmm5[13],xmm12[13],xmm5[14],xmm12[14],xmm5[15],xmm12[15]
-; SSSE3-NEXT:    movdqa %xmm0, %xmm6
-; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
-; SSSE3-NEXT:    pmullw %xmm5, %xmm6
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3],xmm4[4],xmm12[4],xmm4[5],xmm12[5],xmm4[6],xmm12[6],xmm4[7],xmm12[7]
-; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSSE3-NEXT:    pmullw %xmm4, %xmm0
-; SSSE3-NEXT:    psrlw $8, %xmm6
+; SSSE3-NEXT:    packuswb %xmm13, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm10
 ; SSSE3-NEXT:    psrlw $8, %xmm0
-; SSSE3-NEXT:    packuswb %xmm6, %xmm0
-; SSSE3-NEXT:    pcmpeqb %xmm12, %xmm3
-; SSSE3-NEXT:    pcmpeqb %xmm12, %xmm2
-; SSSE3-NEXT:    pcmpeqb %xmm12, %xmm1
-; SSSE3-NEXT:    pcmpeqb %xmm12, %xmm0
+; SSSE3-NEXT:    packuswb %xmm10, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm9, %xmm3
+; SSSE3-NEXT:    pcmpeqb %xmm9, %xmm2
+; SSSE3-NEXT:    pcmpeqb %xmm9, %xmm1
+; SSSE3-NEXT:    pcmpeqb %xmm9, %xmm0
 ; SSSE3-NEXT:    pcmpeqd %xmm4, %xmm4
 ; SSSE3-NEXT:    pxor %xmm4, %xmm3
 ; SSSE3-NEXT:    pxor %xmm4, %xmm2
 ; SSSE3-NEXT:    pxor %xmm4, %xmm1
 ; SSSE3-NEXT:    pxor %xmm4, %xmm0
-; SSSE3-NEXT:    movdqa %xmm11, 48(%rsi)
-; SSSE3-NEXT:    movdqa %xmm10, 32(%rsi)
-; SSSE3-NEXT:    movdqa %xmm9, 16(%rsi)
+; SSSE3-NEXT:    movdqa %xmm8, 48(%rsi)
+; SSSE3-NEXT:    movdqa %xmm14, 32(%rsi)
+; SSSE3-NEXT:    movdqa %xmm12, 16(%rsi)
 ; SSSE3-NEXT:    movdqa %xmm3, %xmm4
 ; SSSE3-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSSE3-NEXT:    movdqa %xmm8, (%rsi)
+; SSSE3-NEXT:    movdqa %xmm11, (%rsi)
 ; SSSE3-NEXT:    movdqa %xmm3, %xmm5
 ; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3]
 ; SSSE3-NEXT:    pslld $31, %xmm3
 ; SSSE3-NEXT:    psrad $31, %xmm3
 ; SSSE3-NEXT:    movdqa %xmm3, 224(%rdi)
 ; SSSE3-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm5

 ; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
 ; SSSE3-NEXT:    pslld $31, %xmm0
 ; SSSE3-NEXT:    psrad $31, %xmm0
 ; SSSE3-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: umulo_v64i8:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa %xmm5, %xmm11
-; SSE41-NEXT:    movdqa %xmm4, %xmm13
-; SSE41-NEXT:    movdqa %xmm3, %xmm4
-; SSE41-NEXT:    movdqa %xmm0, %xmm5
 ; SSE41-NEXT:    movq %rdi, %rax
-; SSE41-NEXT:    movdqa %xmm13, %xmm8
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm8, %xmm0
-; SSE41-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
-; SSE41-NEXT:    pand %xmm10, %xmm0
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm9 = xmm13[0],zero,xmm13[1],zero,xmm13[2],zero,xmm13[3],zero,xmm13[4],zero,xmm13[5],zero,xmm13[6],zero,xmm13[7],zero
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm8 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; SSE41-NEXT:    pmullw %xmm9, %xmm8
-; SSE41-NEXT:    movdqa %xmm8, %xmm3
-; SSE41-NEXT:    pand %xmm10, %xmm3
-; SSE41-NEXT:    packuswb %xmm0, %xmm3
-; SSE41-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE41-NEXT:    movdqa %xmm11, %xmm9
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm9, %xmm0
-; SSE41-NEXT:    pand %xmm10, %xmm0
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm12 = xmm11[0],zero,xmm11[1],zero,xmm11[2],zero,xmm11[3],zero,xmm11[4],zero,xmm11[5],zero,xmm11[6],zero,xmm11[7],zero
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm9 = 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:    pmullw %xmm12, %xmm9
-; SSE41-NEXT:    movdqa %xmm9, %xmm3
-; SSE41-NEXT:    pand %xmm10, %xmm3
-; SSE41-NEXT:    packuswb %xmm0, %xmm3
-; SSE41-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE41-NEXT:    movdqa %xmm6, %xmm12
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm2, %xmm0
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm12, %xmm0
-; SSE41-NEXT:    pand %xmm10, %xmm0
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm14 = 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:    pmovzxbw {{.*#+}} xmm12 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; SSE41-NEXT:    pmullw %xmm14, %xmm12
-; SSE41-NEXT:    movdqa %xmm12, %xmm15
-; SSE41-NEXT:    pand %xmm10, %xmm15
-; SSE41-NEXT:    packuswb %xmm0, %xmm15
-; SSE41-NEXT:    movdqa %xmm7, %xmm0
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    movdqa %xmm4, %xmm3
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE41-NEXT:    pmullw %xmm0, %xmm3
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm0 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm14 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero
-; SSE41-NEXT:    pmullw %xmm0, %xmm14
-; SSE41-NEXT:    pand %xmm10, %xmm3
-; SSE41-NEXT:    pand %xmm14, %xmm10
-; SSE41-NEXT:    packuswb %xmm3, %xmm10
-; SSE41-NEXT:    pxor %xmm0, %xmm0
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm0[8],xmm7[9],xmm0[9],xmm7[10],xmm0[10],xmm7[11],xmm0[11],xmm7[12],xmm0[12],xmm7[13],xmm0[13],xmm7[14],xmm0[14],xmm7[15],xmm0[15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm0[8],xmm4[9],xmm0[9],xmm4[10],xmm0[10],xmm4[11],xmm0[11],xmm4[12],xmm0[12],xmm4[13],xmm0[13],xmm4[14],xmm0[14],xmm4[15],xmm0[15]
-; SSE41-NEXT:    pmullw %xmm7, %xmm4
-; SSE41-NEXT:    psrlw $8, %xmm4
-; SSE41-NEXT:    psrlw $8, %xmm14
-; SSE41-NEXT:    packuswb %xmm4, %xmm14
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
+; SSE41-NEXT:    pxor %xmm13, %xmm13
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm10 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero
+; SSE41-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm13[8],xmm4[9],xmm13[9],xmm4[10],xmm13[10],xmm4[11],xmm13[11],xmm4[12],xmm13[12],xmm4[13],xmm13[13],xmm4[14],xmm13[14],xmm4[15],xmm13[15]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm8 = 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],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE41-NEXT:    pmullw %xmm4, %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255]
+; SSE41-NEXT:    movdqa %xmm0, %xmm4
+; SSE41-NEXT:    pand %xmm9, %xmm4
+; SSE41-NEXT:    pmullw %xmm10, %xmm8
+; SSE41-NEXT:    movdqa %xmm8, %xmm10
+; SSE41-NEXT:    pand %xmm9, %xmm10
+; SSE41-NEXT:    packuswb %xmm4, %xmm10
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm11 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
+; SSE41-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm13[8],xmm5[9],xmm13[9],xmm5[10],xmm13[10],xmm5[11],xmm13[11],xmm5[12],xmm13[12],xmm5[13],xmm13[13],xmm5[14],xmm13[14],xmm5[15],xmm13[15]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm4 = 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],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
+; SSE41-NEXT:    pmullw %xmm5, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, %xmm5
+; SSE41-NEXT:    pand %xmm9, %xmm5
+; SSE41-NEXT:    pmullw %xmm11, %xmm4
+; SSE41-NEXT:    movdqa %xmm4, %xmm11
+; SSE41-NEXT:    pand %xmm9, %xmm11
+; SSE41-NEXT:    packuswb %xmm5, %xmm11
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm12 = 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],xmm13[8],xmm6[9],xmm13[9],xmm6[10],xmm13[10],xmm6[11],xmm13[11],xmm6[12],xmm13[12],xmm6[13],xmm13[13],xmm6[14],xmm13[14],xmm6[15],xmm13[15]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm5 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; SSE41-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE41-NEXT:    pmullw %xmm6, %xmm2
+; SSE41-NEXT:    movdqa %xmm2, %xmm6
+; SSE41-NEXT:    pand %xmm9, %xmm6
+; SSE41-NEXT:    pmullw %xmm12, %xmm5
+; SSE41-NEXT:    movdqa %xmm5, %xmm12
+; SSE41-NEXT:    pand %xmm9, %xmm12
+; SSE41-NEXT:    packuswb %xmm6, %xmm12
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm14 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero,xmm7[4],zero,xmm7[5],zero,xmm7[6],zero,xmm7[7],zero
+; SSE41-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm13[8],xmm7[9],xmm13[9],xmm7[10],xmm13[10],xmm7[11],xmm13[11],xmm7[12],xmm13[12],xmm7[13],xmm13[13],xmm7[14],xmm13[14],xmm7[15],xmm13[15]
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm6 = 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],xmm13[8],xmm3[9],xmm13[9],xmm3[10],xmm13[10],xmm3[11],xmm13[11],xmm3[12],xmm13[12],xmm3[13],xmm13[13],xmm3[14],xmm13[14],xmm3[15],xmm13[15]
+; SSE41-NEXT:    pmullw %xmm7, %xmm3
+; SSE41-NEXT:    pmullw %xmm14, %xmm6
+; SSE41-NEXT:    movdqa %xmm3, %xmm7
+; SSE41-NEXT:    pand %xmm9, %xmm7
+; SSE41-NEXT:    pand %xmm6, %xmm9
+; SSE41-NEXT:    packuswb %xmm7, %xmm9
+; SSE41-NEXT:    psrlw $8, %xmm3
+; SSE41-NEXT:    psrlw $8, %xmm6
+; SSE41-NEXT:    packuswb %xmm3, %xmm6
 ; SSE41-NEXT:    psrlw $8, %xmm2
-; SSE41-NEXT:    psrlw $8, %xmm12
-; SSE41-NEXT:    packuswb %xmm2, %xmm12
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm0[8],xmm11[9],xmm0[9],xmm11[10],xmm0[10],xmm11[11],xmm0[11],xmm11[12],xmm0[12],xmm11[13],xmm0[13],xmm11[14],xmm0[14],xmm11[15],xmm0[15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm1 = 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]
-; SSE41-NEXT:    pmullw %xmm11, %xmm1
-; SSE41-NEXT:    psrlw $8, %xmm1
-; SSE41-NEXT:    psrlw $8, %xmm9
-; SSE41-NEXT:    packuswb %xmm1, %xmm9
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm0[8],xmm13[9],xmm0[9],xmm13[10],xmm0[10],xmm13[11],xmm0[11],xmm13[12],xmm0[12],xmm13[13],xmm0[13],xmm13[14],xmm0[14],xmm13[15],xmm0[15]
-; SSE41-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm0[8],xmm5[9],xmm0[9],xmm5[10],xmm0[10],xmm5[11],xmm0[11],xmm5[12],xmm0[12],xmm5[13],xmm0[13],xmm5[14],xmm0[14],xmm5[15],xmm0[15]
-; SSE41-NEXT:    pmullw %xmm13, %xmm5
 ; SSE41-NEXT:    psrlw $8, %xmm5
+; SSE41-NEXT:    packuswb %xmm2, %xmm5
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm4
+; SSE41-NEXT:    packuswb %xmm1, %xmm4
+; SSE41-NEXT:    psrlw $8, %xmm0
 ; SSE41-NEXT:    psrlw $8, %xmm8
-; SSE41-NEXT:    packuswb %xmm5, %xmm8
-; SSE41-NEXT:    pcmpeqb %xmm0, %xmm14
-; SSE41-NEXT:    pcmpeqb %xmm0, %xmm12
-; SSE41-NEXT:    pcmpeqb %xmm0, %xmm9
-; SSE41-NEXT:    pcmpeqb %xmm0, %xmm8
+; SSE41-NEXT:    packuswb %xmm0, %xmm8
+; SSE41-NEXT:    pcmpeqb %xmm13, %xmm6
+; SSE41-NEXT:    pcmpeqb %xmm13, %xmm5
+; SSE41-NEXT:    pcmpeqb %xmm13, %xmm4
+; SSE41-NEXT:    pcmpeqb %xmm13, %xmm8
 ; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
-; SSE41-NEXT:    pxor %xmm0, %xmm14
-; SSE41-NEXT:    pxor %xmm0, %xmm12
-; SSE41-NEXT:    pxor %xmm0, %xmm9
+; SSE41-NEXT:    pxor %xmm0, %xmm6
+; SSE41-NEXT:    pxor %xmm0, %xmm5
+; SSE41-NEXT:    pxor %xmm0, %xmm4
 ; SSE41-NEXT:    pxor %xmm0, %xmm8
-; SSE41-NEXT:    movdqa %xmm10, 48(%rsi)
-; SSE41-NEXT:    movdqa %xmm15, 32(%rsi)
-; SSE41-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE41-NEXT:    movaps %xmm0, 16(%rsi)
-; SSE41-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE41-NEXT:    movaps %xmm0, (%rsi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm14[0],zero,zero,zero,xmm14[1],zero,zero,zero,xmm14[2],zero,zero,zero,xmm14[3],zero,zero,zero
+; SSE41-NEXT:    movdqa %xmm9, 48(%rsi)
+; SSE41-NEXT:    movdqa %xmm12, 32(%rsi)
+; SSE41-NEXT:    movdqa %xmm11, 16(%rsi)
+; SSE41-NEXT:    movdqa %xmm10, (%rsi)
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm6[0],zero,zero,zero,xmm6[1],zero,zero,zero,xmm6[2],zero,zero,zero,xmm6[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 192(%rdi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm12[0],zero,zero,zero,xmm12[1],zero,zero,zero,xmm12[2],zero,zero,zero,xmm12[3],zero,zero,zero
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 128(%rdi)
-; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm9[0],zero,zero,zero,xmm9[1],zero,zero,zero,xmm9[2],zero,zero,zero,xmm9[3],zero,zero,zero
+; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm4[0],zero,zero,zero,xmm4[1],zero,zero,zero,xmm4[2],zero,zero,zero,xmm4[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 64(%rdi)
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm8[0],zero,zero,zero,xmm8[1],zero,zero,zero,xmm8[2],zero,zero,zero,xmm8[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, (%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 224(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 240(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 208(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 160(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 176(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 144(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 112(%rdi)
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 80(%rdi)
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 32(%rdi)
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 48(%rdi)
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
 ; SSE41-NEXT:    pmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    psrad $31, %xmm0
 ; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: umulo_v64i8:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vmovdqa %ymm2, %ymm10
-; AVX1-NEXT:    vmovdqa %ymm0, %ymm2
 ; AVX1-NEXT:    movq %rdi, %rax
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT:    vpand %xmm7, %xmm4, %xmm4
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm5 = xmm10[0],zero,xmm10[1],zero,xmm10[2],zero,xmm10[3],zero,xmm10[4],zero,xmm10[5],zero,xmm10[6],zero,xmm10[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = 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 %xmm5, %xmm6, %xmm12
-; AVX1-NEXT:    vpand %xmm7, %xmm12, %xmm5
-; AVX1-NEXT:    vpackuswb %xmm4, %xmm5, %xmm0
-; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vextractf128 $1, %ymm10, %xmm13
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm13[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm14
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm14[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpand %xmm7, %xmm4, %xmm5
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm13[0],zero,xmm13[1],zero,xmm13[2],zero,xmm13[3],zero,xmm13[4],zero,xmm13[5],zero,xmm13[6],zero,xmm13[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = xmm14[0],zero,xmm14[1],zero,xmm14[2],zero,xmm14[3],zero,xmm14[4],zero,xmm14[5],zero,xmm14[6],zero,xmm14[7],zero
-; AVX1-NEXT:    vpmullw %xmm4, %xmm6, %xmm15
-; AVX1-NEXT:    vpand %xmm7, %xmm15, %xmm6
-; AVX1-NEXT:    vpackuswb %xmm5, %xmm6, %xmm0
-; AVX1-NEXT:    vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm5, %xmm6, %xmm5
-; AVX1-NEXT:    vpand %xmm7, %xmm5, %xmm5
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm4 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX1-NEXT:    vpmullw %xmm6, %xmm4, %xmm0
-; AVX1-NEXT:    vpand %xmm7, %xmm0, %xmm6
-; AVX1-NEXT:    vpackuswb %xmm5, %xmm6, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm5
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm5[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm11
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm8 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; AVX1-NEXT:    vpmullw %xmm6, %xmm8, %xmm8
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm6 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm9 = xmm11[0],zero,xmm11[1],zero,xmm11[2],zero,xmm11[3],zero,xmm11[4],zero,xmm11[5],zero,xmm11[6],zero,xmm11[7],zero
-; AVX1-NEXT:    vpmullw %xmm6, %xmm9, %xmm6
-; AVX1-NEXT:    vpand %xmm7, %xmm8, %xmm8
-; AVX1-NEXT:    vpand %xmm7, %xmm6, %xmm7
-; AVX1-NEXT:    vpackuswb %xmm8, %xmm7, %xmm8
-; AVX1-NEXT:    vpxor %xmm7, %xmm7, %xmm7
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm7[8],xmm5[9],xmm7[9],xmm5[10],xmm7[10],xmm5[11],xmm7[11],xmm5[12],xmm7[12],xmm5[13],xmm7[13],xmm5[14],xmm7[14],xmm5[15],xmm7[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm11[8],xmm7[8],xmm11[9],xmm7[9],xmm11[10],xmm7[10],xmm11[11],xmm7[11],xmm11[12],xmm7[12],xmm11[13],xmm7[13],xmm11[14],xmm7[14],xmm11[15],xmm7[15]
-; AVX1-NEXT:    vpmullw %xmm5, %xmm4, %xmm4
-; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
-; AVX1-NEXT:    vpsrlw $8, %xmm6, %xmm5
-; AVX1-NEXT:    vpackuswb %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm3[8],xmm7[8],xmm3[9],xmm7[9],xmm3[10],xmm7[10],xmm3[11],xmm7[11],xmm3[12],xmm7[12],xmm3[13],xmm7[13],xmm3[14],xmm7[14],xmm3[15],xmm7[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
-; AVX1-NEXT:    vpmullw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrlw $8, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm13[8],xmm7[8],xmm13[9],xmm7[9],xmm13[10],xmm7[10],xmm13[11],xmm7[11],xmm13[12],xmm7[12],xmm13[13],xmm7[13],xmm13[14],xmm7[14],xmm13[15],xmm7[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm14[8],xmm7[8],xmm14[9],xmm7[9],xmm14[10],xmm7[10],xmm14[11],xmm7[11],xmm14[12],xmm7[12],xmm14[13],xmm7[13],xmm14[14],xmm7[14],xmm14[15],xmm7[15]
-; AVX1-NEXT:    vpmullw %xmm1, %xmm3, %xmm1
-; AVX1-NEXT:    vpsrlw $8, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrlw $8, %xmm15, %xmm3
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm3, %xmm1
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm10[8],xmm7[8],xmm10[9],xmm7[9],xmm10[10],xmm7[10],xmm10[11],xmm7[11],xmm10[12],xmm7[12],xmm10[13],xmm7[13],xmm10[14],xmm7[14],xmm10[15],xmm7[15]
-; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; AVX1-NEXT:    vpmullw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpxor %xmm5, %xmm5, %xmm5
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm6 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
+; AVX1-NEXT:    vpmullw %xmm4, %xmm6, %xmm9
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT:    vpand %xmm6, %xmm9, %xmm8
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm7 = 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:    vpmovzxbw {{.*#+}} xmm4 = 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 %xmm7, %xmm4, %xmm11
+; AVX1-NEXT:    vpand %xmm6, %xmm11, %xmm4
+; AVX1-NEXT:    vpackuswb %xmm8, %xmm4, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm2
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm7 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
+; AVX1-NEXT:    vpmullw %xmm4, %xmm7, %xmm12
+; AVX1-NEXT:    vpand %xmm6, %xmm12, %xmm7
+; 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:    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 %xmm2, %xmm0, %xmm13
+; AVX1-NEXT:    vpand %xmm6, %xmm13, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm7, %xmm2, %xmm10
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm7 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
+; AVX1-NEXT:    vpmullw %xmm2, %xmm7, %xmm7
+; AVX1-NEXT:    vpand %xmm6, %xmm7, %xmm2
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} 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
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVX1-NEXT:    vpmullw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm6, %xmm0, %xmm4
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm4, %xmm14
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm3
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm1
+; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
+; AVX1-NEXT:    vpmullw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVX1-NEXT:    vpmullw %xmm3, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm6, %xmm2, %xmm1
+; AVX1-NEXT:    vpand %xmm6, %xmm3, %xmm4
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm4, %xmm15
 ; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
-; AVX1-NEXT:    vpsrlw $8, %xmm12, %xmm3
+; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm3
 ; AVX1-NEXT:    vpackuswb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpeqb %xmm7, %xmm4, %xmm3
-; AVX1-NEXT:    vpcmpeqb %xmm7, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm7, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpeqb %xmm7, %xmm2, %xmm4
-; AVX1-NEXT:    vpcmpeqd %xmm5, %xmm5, %xmm5
-; AVX1-NEXT:    vpxor %xmm5, %xmm3, %xmm3
-; AVX1-NEXT:    vpxor %xmm5, %xmm0, %xmm2
-; AVX1-NEXT:    vpxor %xmm5, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm5, %xmm4, %xmm0
-; AVX1-NEXT:    vmovdqa %xmm8, 48(%rsi)
-; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-NEXT:    vmovaps %xmm4, 32(%rsi)
-; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-NEXT:    vmovaps %xmm4, 16(%rsi)
-; AVX1-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-NEXT:    vmovaps %xmm4, (%rsi)
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, 192(%rdi)
-; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, 128(%rdi)
-; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, 64(%rdi)
-; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, (%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, 224(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[3,3,3,3]
-; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm4
-; AVX1-NEXT:    vmovdqa %xmm4, 240(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
-; AVX1-NEXT:    vmovdqa %xmm3, 208(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
-; AVX1-NEXT:    vmovdqa %xmm3, 160(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[3,3,3,3]
-; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm3
-; AVX1-NEXT:    vmovdqa %xmm3, 176(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
-; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
-; AVX1-NEXT:    vmovdqa %xmm2, 144(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
-; AVX1-NEXT:    vmovdqa %xmm2, 96(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
-; AVX1-NEXT:    vpmovsxbd %xmm2, %xmm2
-; AVX1-NEXT:    vmovdqa %xmm2, 112(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa %xmm1, 80(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa %xmm1, 32(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX1-NEXT:    vpmovsxbd %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa %xmm1, 48(%rdi)
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; AVX1-NEXT:    vpsrlw $8, %xmm7, %xmm3
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm12, %xmm3
+; AVX1-NEXT:    vpsrlw $8, %xmm13, %xmm4
+; AVX1-NEXT:    vpackuswb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT:    vpsrlw $8, %xmm9, %xmm4
+; AVX1-NEXT:    vpsrlw $8, %xmm11, %xmm6
+; AVX1-NEXT:    vpackuswb %xmm4, %xmm6, %xmm4
+; AVX1-NEXT:    vpcmpeqb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpeqb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm5, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm5, %xmm4, %xmm7
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm2, %xmm6
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpxor %xmm1, %xmm3, %xmm5
+; AVX1-NEXT:    vpxor %xmm1, %xmm7, %xmm3
+; AVX1-NEXT:    vmovdqa %xmm15, 48(%rsi)
+; AVX1-NEXT:    vmovdqa %xmm14, 32(%rsi)
+; AVX1-NEXT:    vmovdqa %xmm10, 16(%rsi)
+; AVX1-NEXT:    vmovdqa %xmm8, (%rsi)
+; AVX1-NEXT:    vpmovsxbd %xmm6, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 192(%rdi)
+; AVX1-NEXT:    vpmovsxbd %xmm4, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 128(%rdi)
+; AVX1-NEXT:    vpmovsxbd %xmm5, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 64(%rdi)
+; AVX1-NEXT:    vpmovsxbd %xmm3, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, (%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 224(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 240(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 208(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 160(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 176(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 144(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 96(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 112(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 80(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 32(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
+; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa %xmm0, 48(%rdi)
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
 ; AVX1-NEXT:    vpmovsxbd %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqa %xmm0, 16(%rdi)
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: umulo_v64i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    movq %rdi, %rax
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm4 = ymm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpmullw %ymm4, %ymm5, %ymm4
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT:    vpand %ymm5, %ymm4, %ymm4
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm6 = ymm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm7 = ymm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpmullw %ymm6, %ymm7, %ymm6
-; AVX2-NEXT:    vpand %ymm5, %ymm6, %ymm6
-; AVX2-NEXT:    vpackuswb %ymm4, %ymm6, %ymm9
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm6 = ymm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm7 = ymm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31]
-; AVX2-NEXT:    vpmullw %ymm6, %ymm7, %ymm6
-; AVX2-NEXT:    vpand %ymm5, %ymm6, %ymm6
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm7 = ymm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm8 = ymm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23]
+; AVX2-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm5 = ymm2[8],ymm4[8],ymm2[9],ymm4[9],ymm2[10],ymm4[10],ymm2[11],ymm4[11],ymm2[12],ymm4[12],ymm2[13],ymm4[13],ymm2[14],ymm4[14],ymm2[15],ymm4[15],ymm2[24],ymm4[24],ymm2[25],ymm4[25],ymm2[26],ymm4[26],ymm2[27],ymm4[27],ymm2[28],ymm4[28],ymm2[29],ymm4[29],ymm2[30],ymm4[30],ymm2[31],ymm4[31]
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm6 = ymm0[8],ymm4[8],ymm0[9],ymm4[9],ymm0[10],ymm4[10],ymm0[11],ymm4[11],ymm0[12],ymm4[12],ymm0[13],ymm4[13],ymm0[14],ymm4[14],ymm0[15],ymm4[15],ymm0[24],ymm4[24],ymm0[25],ymm4[25],ymm0[26],ymm4[26],ymm0[27],ymm4[27],ymm0[28],ymm4[28],ymm0[29],ymm4[29],ymm0[30],ymm4[30],ymm0[31],ymm4[31]
+; AVX2-NEXT:    vpmullw %ymm5, %ymm6, %ymm5
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-NEXT:    vpand %ymm6, %ymm5, %ymm7
+; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[4],ymm4[4],ymm2[5],ymm4[5],ymm2[6],ymm4[6],ymm2[7],ymm4[7],ymm2[16],ymm4[16],ymm2[17],ymm4[17],ymm2[18],ymm4[18],ymm2[19],ymm4[19],ymm2[20],ymm4[20],ymm2[21],ymm4[21],ymm2[22],ymm4[22],ymm2[23],ymm4[23]
+; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[2],ymm4[2],ymm0[3],ymm4[3],ymm0[4],ymm4[4],ymm0[5],ymm4[5],ymm0[6],ymm4[6],ymm0[7],ymm4[7],ymm0[16],ymm4[16],ymm0[17],ymm4[17],ymm0[18],ymm4[18],ymm0[19],ymm4[19],ymm0[20],ymm4[20],ymm0[21],ymm4[21],ymm0[22],ymm4[22],ymm0[23],ymm4[23]
+; AVX2-NEXT:    vpmullw %ymm2, %ymm0, %ymm2
+; AVX2-NEXT:    vpand %ymm6, %ymm2, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm7, %ymm0, %ymm9
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm7 = ymm3[8],ymm4[8],ymm3[9],ymm4[9],ymm3[10],ymm4[10],ymm3[11],ymm4[11],ymm3[12],ymm4[12],ymm3[13],ymm4[13],ymm3[14],ymm4[14],ymm3[15],ymm4[15],ymm3[24],ymm4[24],ymm3[25],ymm4[25],ymm3[26],ymm4[26],ymm3[27],ymm4[27],ymm3[28],ymm4[28],ymm3[29],ymm4[29],ymm3[30],ymm4[30],ymm3[31],ymm4[31]
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm8 = ymm1[8],ymm4[8],ymm1[9],ymm4[9],ymm1[10],ymm4[10],ymm1[11],ymm4[11],ymm1[12],ymm4[12],ymm1[13],ymm4[13],ymm1[14],ymm4[14],ymm1[15],ymm4[15],ymm1[24],ymm4[24],ymm1[25],ymm4[25],ymm1[26],ymm4[26],ymm1[27],ymm4[27],ymm1[28],ymm4[28],ymm1[29],ymm4[29],ymm1[30],ymm4[30],ymm1[31],ymm4[31]
 ; AVX2-NEXT:    vpmullw %ymm7, %ymm8, %ymm7
-; AVX2-NEXT:    vpand %ymm5, %ymm7, %ymm5
-; AVX2-NEXT:    vpackuswb %ymm6, %ymm5, %ymm10
-; AVX2-NEXT:    vpxor %xmm6, %xmm6, %xmm6
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm7 = ymm3[8],ymm6[8],ymm3[9],ymm6[9],ymm3[10],ymm6[10],ymm3[11],ymm6[11],ymm3[12],ymm6[12],ymm3[13],ymm6[13],ymm3[14],ymm6[14],ymm3[15],ymm6[15],ymm3[24],ymm6[24],ymm3[25],ymm6[25],ymm3[26],ymm6[26],ymm3[27],ymm6[27],ymm3[28],ymm6[28],ymm3[29],ymm6[29],ymm3[30],ymm6[30],ymm3[31],ymm6[31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm8 = ymm1[8],ymm6[8],ymm1[9],ymm6[9],ymm1[10],ymm6[10],ymm1[11],ymm6[11],ymm1[12],ymm6[12],ymm1[13],ymm6[13],ymm1[14],ymm6[14],ymm1[15],ymm6[15],ymm1[24],ymm6[24],ymm1[25],ymm6[25],ymm1[26],ymm6[26],ymm1[27],ymm6[27],ymm1[28],ymm6[28],ymm1[29],ymm6[29],ymm1[30],ymm6[30],ymm1[31],ymm6[31]
-; AVX2-NEXT:    vpmullw %ymm7, %ymm8, %ymm7
-; AVX2-NEXT:    vpsrlw $8, %ymm7, %ymm7
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[1],ymm6[1],ymm3[2],ymm6[2],ymm3[3],ymm6[3],ymm3[4],ymm6[4],ymm3[5],ymm6[5],ymm3[6],ymm6[6],ymm3[7],ymm6[7],ymm3[16],ymm6[16],ymm3[17],ymm6[17],ymm3[18],ymm6[18],ymm3[19],ymm6[19],ymm3[20],ymm6[20],ymm3[21],ymm6[21],ymm3[22],ymm6[22],ymm3[23],ymm6[23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm6[0],ymm1[1],ymm6[1],ymm1[2],ymm6[2],ymm1[3],ymm6[3],ymm1[4],ymm6[4],ymm1[5],ymm6[5],ymm1[6],ymm6[6],ymm1[7],ymm6[7],ymm1[16],ymm6[16],ymm1[17],ymm6[17],ymm1[18],ymm6[18],ymm1[19],ymm6[19],ymm1[20],ymm6[20],ymm1[21],ymm6[21],ymm1[22],ymm6[22],ymm1[23],ymm6[23]
+; AVX2-NEXT:    vpand %ymm6, %ymm7, %ymm8
+; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[4],ymm4[4],ymm3[5],ymm4[5],ymm3[6],ymm4[6],ymm3[7],ymm4[7],ymm3[16],ymm4[16],ymm3[17],ymm4[17],ymm3[18],ymm4[18],ymm3[19],ymm4[19],ymm3[20],ymm4[20],ymm3[21],ymm4[21],ymm3[22],ymm4[22],ymm3[23],ymm4[23]
+; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[4],ymm4[4],ymm1[5],ymm4[5],ymm1[6],ymm4[6],ymm1[7],ymm4[7],ymm1[16],ymm4[16],ymm1[17],ymm4[17],ymm1[18],ymm4[18],ymm1[19],ymm4[19],ymm1[20],ymm4[20],ymm1[21],ymm4[21],ymm1[22],ymm4[22],ymm1[23],ymm4[23]
 ; AVX2-NEXT:    vpmullw %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vpand %ymm6, %ymm1, %ymm3
+; AVX2-NEXT:    vpackuswb %ymm8, %ymm3, %ymm8
+; AVX2-NEXT:    vpsrlw $8, %ymm7, %ymm6
 ; AVX2-NEXT:    vpsrlw $8, %ymm1, %ymm1
-; AVX2-NEXT:    vpackuswb %ymm7, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqb %ymm6, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpeqd %ymm3, %ymm3, %ymm3
-; AVX2-NEXT:    vpxor %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm7 = ymm2[8],ymm6[8],ymm2[9],ymm6[9],ymm2[10],ymm6[10],ymm2[11],ymm6[11],ymm2[12],ymm6[12],ymm2[13],ymm6[13],ymm2[14],ymm6[14],ymm2[15],ymm6[15],ymm2[24],ymm6[24],ymm2[25],ymm6[25],ymm2[26],ymm6[26],ymm2[27],ymm6[27],ymm2[28],ymm6[28],ymm2[29],ymm6[29],ymm2[30],ymm6[30],ymm2[31],ymm6[31]
-; AVX2-NEXT:    vpunpckhbw {{.*#+}} ymm8 = ymm0[8],ymm6[8],ymm0[9],ymm6[9],ymm0[10],ymm6[10],ymm0[11],ymm6[11],ymm0[12],ymm6[12],ymm0[13],ymm6[13],ymm0[14],ymm6[14],ymm0[15],ymm6[15],ymm0[24],ymm6[24],ymm0[25],ymm6[25],ymm0[26],ymm6[26],ymm0[27],ymm6[27],ymm0[28],ymm6[28],ymm0[29],ymm6[29],ymm0[30],ymm6[30],ymm0[31],ymm6[31]
-; AVX2-NEXT:    vpmullw %ymm7, %ymm8, %ymm7
-; AVX2-NEXT:    vpsrlw $8, %ymm7, %ymm7
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm2 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[2],ymm6[2],ymm2[3],ymm6[3],ymm2[4],ymm6[4],ymm2[5],ymm6[5],ymm2[6],ymm6[6],ymm2[7],ymm6[7],ymm2[16],ymm6[16],ymm2[17],ymm6[17],ymm2[18],ymm6[18],ymm2[19],ymm6[19],ymm2[20],ymm6[20],ymm2[21],ymm6[21],ymm2[22],ymm6[22],ymm2[23],ymm6[23]
-; AVX2-NEXT:    vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[1],ymm6[1],ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[4],ymm6[4],ymm0[5],ymm6[5],ymm0[6],ymm6[6],ymm0[7],ymm6[7],ymm0[16],ymm6[16],ymm0[17],ymm6[17],ymm0[18],ymm6[18],ymm0[19],ymm6[19],ymm0[20],ymm6[20],ymm0[21],ymm6[21],ymm0[22],ymm6[22],ymm0[23],ymm6[23]
-; AVX2-NEXT:    vpmullw %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
-; AVX2-NEXT:    vpackuswb %ymm7, %ymm0, %ymm0
-; AVX2-NEXT:    vpcmpeqb %ymm6, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %ymm3, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
-; AVX2-NEXT:    vpmovsxbd %xmm2, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
+; AVX2-NEXT:    vpackuswb %ymm6, %ymm1, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm4, %ymm1, %ymm1
+; AVX2-NEXT:    vpcmpeqd %ymm6, %ymm6, %ymm6
+; AVX2-NEXT:    vpxor %ymm6, %ymm1, %ymm1
+; AVX2-NEXT:    vpsrlw $8, %ymm5, %ymm5
+; AVX2-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX2-NEXT:    vpackuswb %ymm5, %ymm2, %ymm2
+; AVX2-NEXT:    vpcmpeqb %ymm4, %ymm2, %ymm2
+; AVX2-NEXT:    vpxor %ymm6, %ymm2, %ymm2
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[2,3,2,3]
+; AVX2-NEXT:    vpmovsxbd %xmm4, %ymm4
+; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm5
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm6 = xmm5[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm6, %ymm6
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm7 = xmm1[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm7, %ymm7
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm4
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[2,3,2,3]
-; AVX2-NEXT:    vpmovsxbd %xmm5, %ymm5
-; AVX2-NEXT:    vpmovsxbd %xmm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm0
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpmovsxbd %xmm3, %ymm3
+; AVX2-NEXT:    vpmovsxbd %xmm2, %ymm2
+; AVX2-NEXT:    vpmovsxbd %xmm5, %ymm5
 ; AVX2-NEXT:    vpmovsxbd %xmm1, %ymm1
-; AVX2-NEXT:    vpmovsxbd %xmm4, %ymm4
-; AVX2-NEXT:    vmovdqa %ymm10, 32(%rsi)
+; AVX2-NEXT:    vpmovsxbd %xmm0, %ymm0
+; AVX2-NEXT:    vmovdqa %ymm8, 32(%rsi)
 ; AVX2-NEXT:    vmovdqa %ymm9, (%rsi)
-; AVX2-NEXT:    vmovdqa %ymm4, 192(%rdi)
+; AVX2-NEXT:    vmovdqa %ymm0, 192(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm1, 128(%rdi)
-; AVX2-NEXT:    vmovdqa %ymm3, 64(%rdi)
-; AVX2-NEXT:    vmovdqa %ymm0, (%rdi)
-; AVX2-NEXT:    vmovdqa %ymm5, 224(%rdi)
+; AVX2-NEXT:    vmovdqa %ymm5, 64(%rdi)
+; AVX2-NEXT:    vmovdqa %ymm2, (%rdi)
+; AVX2-NEXT:    vmovdqa %ymm3, 224(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm7, 160(%rdi)
 ; AVX2-NEXT:    vmovdqa %ymm6, 96(%rdi)
-; AVX2-NEXT:    vmovdqa %ymm2, 32(%rdi)
+; AVX2-NEXT:    vmovdqa %ymm4, 32(%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: umulo_v64i8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vextracti64x4 $1, %zmm1, %ymm3
 ; AVX512F-NEXT:    vextracti128 $1, %ymm3, %xmm2
 ; AVX512F-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero

 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm4 = ymm6[0],zero,ymm6[1],zero,ymm6[2],zero,ymm6[3],zero,ymm6[4],zero,ymm6[5],zero,ymm6[6],zero,ymm6[7],zero,ymm6[8],zero,ymm6[9],zero,ymm6[10],zero,ymm6[11],zero,ymm6[12],zero,ymm6[13],zero,ymm6[14],zero,ymm6[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm4, 16(%rdi)
 ; AVX512F-NEXT:    vpmovzxwd {{.*#+}} zmm4 = ymm7[0],zero,ymm7[1],zero,ymm7[2],zero,ymm7[3],zero,ymm7[4],zero,ymm7[5],zero,ymm7[6],zero,ymm7[7],zero,ymm7[8],zero,ymm7[9],zero,ymm7[10],zero,ymm7[11],zero,ymm7[12],zero,ymm7[13],zero,ymm7[14],zero,ymm7[15],zero
 ; AVX512F-NEXT:    vpmovdb %zmm4, (%rdi)
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512BW-LABEL: umulo_v64i8:
 ; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm2 = zmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
-; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm3 = zmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15,24,24,25,25,26,26,27,27,28,28,29,29,30,30,31,31,40,40,41,41,42,42,43,43,44,44,45,45,46,46,47,47,56,56,57,57,58,58,59,59,60,60,61,61,62,62,63,63]
-; AVX512BW-NEXT:    vpmullw %zmm2, %zmm3, %zmm2
-; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512BW-NEXT:    vpandq %zmm3, %zmm2, %zmm2
-; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm4 = zmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
-; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm5 = zmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,16,16,17,17,18,18,19,19,20,20,21,21,22,22,23,23,32,32,33,33,34,34,35,35,36,36,37,37,38,38,39,39,48,48,49,49,50,50,51,51,52,52,53,53,54,54,55,55]
-; AVX512BW-NEXT:    vpmullw %zmm4, %zmm5, %zmm4
-; AVX512BW-NEXT:    vpandq %zmm3, %zmm4, %zmm3
-; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm3, %zmm4
 ; AVX512BW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm3 = zmm1[8],zmm2[8],zmm1[9],zmm2[9],zmm1[10],zmm2[10],zmm1[11],zmm2[11],zmm1[12],zmm2[12],zmm1[13],zmm2[13],zmm1[14],zmm2[14],zmm1[15],zmm2[15],zmm1[24],zmm2[24],zmm1[25],zmm2[25],zmm1[26],zmm2[26],zmm1[27],zmm2[27],zmm1[28],zmm2[28],zmm1[29],zmm2[29],zmm1[30],zmm2[30],zmm1[31],zmm2[31],zmm1[40],zmm2[40],zmm1[41],zmm2[41],zmm1[42],zmm2[42],zmm1[43],zmm2[43],zmm1[44],zmm2[44],zmm1[45],zmm2[45],zmm1[46],zmm2[46],zmm1[47],zmm2[47],zmm1[56],zmm2[56],zmm1[57],zmm2[57],zmm1[58],zmm2[58],zmm1[59],zmm2[59],zmm1[60],zmm2[60],zmm1[61],zmm2[61],zmm1[62],zmm2[62],zmm1[63],zmm2[63]
-; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm5 = zmm0[8],zmm2[8],zmm0[9],zmm2[9],zmm0[10],zmm2[10],zmm0[11],zmm2[11],zmm0[12],zmm2[12],zmm0[13],zmm2[13],zmm0[14],zmm2[14],zmm0[15],zmm2[15],zmm0[24],zmm2[24],zmm0[25],zmm2[25],zmm0[26],zmm2[26],zmm0[27],zmm2[27],zmm0[28],zmm2[28],zmm0[29],zmm2[29],zmm0[30],zmm2[30],zmm0[31],zmm2[31],zmm0[40],zmm2[40],zmm0[41],zmm2[41],zmm0[42],zmm2[42],zmm0[43],zmm2[43],zmm0[44],zmm2[44],zmm0[45],zmm2[45],zmm0[46],zmm2[46],zmm0[47],zmm2[47],zmm0[56],zmm2[56],zmm0[57],zmm2[57],zmm0[58],zmm2[58],zmm0[59],zmm2[59],zmm0[60],zmm2[60],zmm0[61],zmm2[61],zmm0[62],zmm2[62],zmm0[63],zmm2[63]
-; AVX512BW-NEXT:    vpmullw %zmm3, %zmm5, %zmm3
-; AVX512BW-NEXT:    vpsrlw $8, %zmm3, %zmm3
+; AVX512BW-NEXT:    vpunpckhbw {{.*#+}} zmm4 = zmm0[8],zmm2[8],zmm0[9],zmm2[9],zmm0[10],zmm2[10],zmm0[11],zmm2[11],zmm0[12],zmm2[12],zmm0[13],zmm2[13],zmm0[14],zmm2[14],zmm0[15],zmm2[15],zmm0[24],zmm2[24],zmm0[25],zmm2[25],zmm0[26],zmm2[26],zmm0[27],zmm2[27],zmm0[28],zmm2[28],zmm0[29],zmm2[29],zmm0[30],zmm2[30],zmm0[31],zmm2[31],zmm0[40],zmm2[40],zmm0[41],zmm2[41],zmm0[42],zmm2[42],zmm0[43],zmm2[43],zmm0[44],zmm2[44],zmm0[45],zmm2[45],zmm0[46],zmm2[46],zmm0[47],zmm2[47],zmm0[56],zmm2[56],zmm0[57],zmm2[57],zmm0[58],zmm2[58],zmm0[59],zmm2[59],zmm0[60],zmm2[60],zmm0[61],zmm2[61],zmm0[62],zmm2[62],zmm0[63],zmm2[63]
+; AVX512BW-NEXT:    vpmullw %zmm3, %zmm4, %zmm3
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512BW-NEXT:    vpandq %zmm4, %zmm3, %zmm5
 ; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm1 = zmm1[0],zmm2[0],zmm1[1],zmm2[1],zmm1[2],zmm2[2],zmm1[3],zmm2[3],zmm1[4],zmm2[4],zmm1[5],zmm2[5],zmm1[6],zmm2[6],zmm1[7],zmm2[7],zmm1[16],zmm2[16],zmm1[17],zmm2[17],zmm1[18],zmm2[18],zmm1[19],zmm2[19],zmm1[20],zmm2[20],zmm1[21],zmm2[21],zmm1[22],zmm2[22],zmm1[23],zmm2[23],zmm1[32],zmm2[32],zmm1[33],zmm2[33],zmm1[34],zmm2[34],zmm1[35],zmm2[35],zmm1[36],zmm2[36],zmm1[37],zmm2[37],zmm1[38],zmm2[38],zmm1[39],zmm2[39],zmm1[48],zmm2[48],zmm1[49],zmm2[49],zmm1[50],zmm2[50],zmm1[51],zmm2[51],zmm1[52],zmm2[52],zmm1[53],zmm2[53],zmm1[54],zmm2[54],zmm1[55],zmm2[55]
 ; AVX512BW-NEXT:    vpunpcklbw {{.*#+}} zmm0 = zmm0[0],zmm2[0],zmm0[1],zmm2[1],zmm0[2],zmm2[2],zmm0[3],zmm2[3],zmm0[4],zmm2[4],zmm0[5],zmm2[5],zmm0[6],zmm2[6],zmm0[7],zmm2[7],zmm0[16],zmm2[16],zmm0[17],zmm2[17],zmm0[18],zmm2[18],zmm0[19],zmm2[19],zmm0[20],zmm2[20],zmm0[21],zmm2[21],zmm0[22],zmm2[22],zmm0[23],zmm2[23],zmm0[32],zmm2[32],zmm0[33],zmm2[33],zmm0[34],zmm2[34],zmm0[35],zmm2[35],zmm0[36],zmm2[36],zmm0[37],zmm2[37],zmm0[38],zmm2[38],zmm0[39],zmm2[39],zmm0[48],zmm2[48],zmm0[49],zmm2[49],zmm0[50],zmm2[50],zmm0[51],zmm2[51],zmm0[52],zmm2[52],zmm0[53],zmm2[53],zmm0[54],zmm2[54],zmm0[55],zmm2[55]
 ; AVX512BW-NEXT:    vpmullw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm4, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpackuswb %zmm5, %zmm1, %zmm4
+; AVX512BW-NEXT:    vpsrlw $8, %zmm3, %zmm1
 ; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
-; AVX512BW-NEXT:    vpackuswb %zmm3, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    vptestmb %zmm0, %zmm0, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k2
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k2} {z}
 ; AVX512BW-NEXT:    kshiftrq $32, %k1, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm2, %zmm2, %zmm2 {%k1} {z}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k1
 ; AVX512BW-NEXT:    vpternlogd $255, %zmm3, %zmm3, %zmm3 {%k1} {z}