[X86] Match bitcast of vxi1 to pmovmsk

[zvi]

Summary

Add an early combine to match patterns such as:

(i16 bitcast (v16i1 x))
->
(i16 movmsk (v16i8 sext (v16i1 x)))

This combine needs to happen early enough before
type-legalization scalarizes the result of the setcc.

[zvi]

Rebase

[igorb]

LGTM

[RKSimon]

Why don't you use MOVMSKPD/MOVMSKPS for the 32/64 bit cases and avoid the vector truncation?

[zvi]

In D33311#758420, @RKSimon wrote:

Why don't you use MOVMSKPD/MOVMSKPS for the 32/64 bit cases and avoid the vector truncation?

That's a good point. thanks!

[delena]

lib/Target/X86/X86ISelLowering.cpp

29020 ↗

(On Diff #99418)

Can you use if (VT.isInteger() && DAG.ComputeNumSignBits(N) == EltBits) ?

35127 ↗

(On Diff #99418)

We usually pass DCI to check the phase.

[zvi]

Thanks, Simon, for reminding me that the integer and FP MOVMSK operations are assymetrical, so we get more types covered without having to shuffle.

[zvi]

lib/Target/X86/X86ISelLowering.cpp

29020 ↗

(On Diff #99418)

i didn't understand the question. Can you please explain this point?

[delena]

lib/Target/X86/X86ISelLowering.cpp

29020 ↗

(On Diff #99418)

ISD::SETCC is a private case when you know about all-ones or all-zeroes. DAG.ComputeNumSignBits(N) == EltBits is more general.

[zvi]

lib/Target/X86/X86ISelLowering.cpp

29020 ↗

(On Diff #99418)

That's a good idea. Do you have an example of such a case?

[zvi]

Generalize to catch cases where the setcc operands are illegal types.

[RKSimon]

lib/Target/X86/X86ISelLowering.cpp

29020 ↗

(On Diff #99418)

combineVectorSignBitsTruncation should be a good example - it started with using the 'all bits' of a comparison result and was generalised to use DAG.ComputeNumSignBits later on

[zvi]

Generalize combine to match (ix bitcast (vxi1 t)) where t is not necessarily a setcc.

[RKSimon]

LGTM with one minor - thanks.

lib/Target/X86/X86ISelLowering.cpp

29017 ↗

(On Diff #100664)

Possibly add a TODO about handling pre-AVX2 cases (split with 2 16i8 MOVMSK) - add a SSE/AVX1 test with current codegen as well if you can.

[zvi]

Addressing Simon's comments.

llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

static SDValue XFormVExtractWithShuffleIntoLoad(SDNode *N, SelectionDAG &DAG,
   SDValue Shuffle = (UnaryShuffle) ? DAG.getUNDEF(CurrentVT) : ShuffleOps[1];
   Shuffle = DAG.getVectorShuffle(CurrentVT, dl, ShuffleOps[0], Shuffle,
                                  ShuffleMask);
   Shuffle = DAG.getBitcast(OriginalVT, Shuffle);
   return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, N->getValueType(0), Shuffle,
                      EltNo);
 }
 
+// Try to match patterns such as
+// (i16 bitcast (v16i1 x))
+// ->
+// (i16 movmsk (16i8 sext (v16i1 x)))
+// before the illegal vector is scalarized on subtargets that don't have legal
+// vxi1 types.
+static SDValue combineBitcastvxi1(SelectionDAG &DAG, SDValue BitCast,
+                                  const X86Subtarget &Subtarget) {
+  EVT VT = BitCast.getValueType();
+  SDValue N0 = BitCast.getOperand(0);
+  EVT VecVT = N0->getValueType(0);
+
+  if (!VT.isScalarInteger() || !VecVT.isSimple())
+    return SDValue();
+
+  // With AVX512 vxi1 types are legal and we prefer using k-regs.
+  // MOVMSK is supported in SSE2 or later.
+  if (Subtarget.hasAVX512() || !Subtarget.hasSSE2())
+    return SDValue();
+
+  // There are MOVMSK flavors for types v16i8, v32i8, v4f32, v8f32, v4f64 and
+  // v8f64. So all legal 128-bit and 256-bit vectors are covered except for
+  // v8i16 and v16i16.
+  // For these two cases, we can shuffle the upper element bytes to a
+  // consecutive sequence at the start of the vector and treat the results as
+  // v16i8 or v32i8, and for v61i8 this is the prefferable solution. However,
+  // for v16i16 this is not the case, because the shuffle is expensive, so we
+  // avoid sign-exteding to this type entirely.
+  // For example, t0 := (v8i16 sext(v8i1 x)) needs to be shuffled as:
+  // (v16i8 shuffle <0,2,4,6,8,10,12,14,u,u,...,u> (v16i8 bitcast t0), undef)
+  MVT SExtVT;
+  MVT FPCastVT = MVT::INVALID_SIMPLE_VALUE_TYPE;
+  switch (VecVT.getSimpleVT().SimpleTy) {
+  default:
+    return SDValue();
+  case MVT::v2i1:
+    SExtVT = MVT::v2i64;
+    FPCastVT = MVT::v2f64;
+    break;
+  case MVT::v4i1:
+    SExtVT = MVT::v4i32;
+    FPCastVT = MVT::v4f32;
+    // For cases such as (i4 bitcast (v4i1 setcc v4i64 v1, v2))
+    // sign-extend to a 256-bit operation to avoid truncation.
+    if (N0->getOpcode() == ISD::SETCC &&
+        N0->getOperand(0)->getValueType(0).is256BitVector() &&
+        Subtarget.hasInt256()) {
+      SExtVT = MVT::v4i64;
+      FPCastVT = MVT::v4f64;
+    }
+    break;
+  case MVT::v8i1:
+    SExtVT = MVT::v8i16;
+    // For cases such as (i8 bitcast (v8i1 setcc v8i32 v1, v2)),
+    // sign-extend to a 256-bit operation to match the compare.
+    // If the setcc operand is 128-bit, prefer sign-extending to 128-bit over
+    // 256-bit because the shuffle is cheaper than sign extending the result of
+    // the compare.
+    if (N0->getOpcode() == ISD::SETCC &&
+        N0->getOperand(0)->getValueType(0).is256BitVector() &&
+        Subtarget.hasInt256()) {
+      SExtVT = MVT::v8i32;
+      FPCastVT = MVT::v8f32;
+    }
+    break;
+  case MVT::v16i1:
+    SExtVT = MVT::v16i8;
+    // For the case (i16 bitcast (v16i1 setcc v16i16 v1, v2)),
+    // it is not profitable to sign-extend to 256-bit because this will
+    // require an extra cross-lane shuffle which is more exprensive than
+    // truncating the result of the compare to 128-bits.
+    break;
+  case MVT::v32i1:
+    // TODO: Handle pre-AVX2 cases by splitting to two v16i1's.
+    if (!Subtarget.hasInt256())
+      return SDValue();
+    SExtVT = MVT::v32i8;
+    break;
+  };
+
+  SDLoc DL(BitCast);
+  SDValue V = DAG.getSExtOrTrunc(N0, DL, SExtVT);
+  if (SExtVT == MVT::v8i16) {
+    V = DAG.getBitcast(MVT::v16i8, V);
+    V = DAG.getVectorShuffle(
+        MVT::v16i8, DL, V, DAG.getUNDEF(MVT::v16i8),
+        {0, 2, 4, 6, 8, 10, 12, 14, -1, -1, -1, -1, -1, -1, -1, -1});
+  } else
+    assert(SExtVT.getScalarType() != MVT::i16 &&
+           "Vectors of i16 must be shuffled");
+  if (FPCastVT != MVT::INVALID_SIMPLE_VALUE_TYPE)
+    V = DAG.getBitcast(FPCastVT, V);
+  V = DAG.getNode(X86ISD::MOVMSK, DL, MVT::i32, V);
+  return DAG.getZExtOrTrunc(V, DL, VT);
+}
+
 static SDValue combineBitcast(SDNode *N, SelectionDAG &DAG,
+                              TargetLowering::DAGCombinerInfo &DCI,
                               const X86Subtarget &Subtarget) {
   SDValue N0 = N->getOperand(0);
   EVT VT = N->getValueType(0);
   EVT SrcVT = N0.getValueType();
 
+  // Try to match patterns such as
+  // (i16 bitcast (v16i1 x))
+  // ->
+  // (i16 movmsk (16i8 sext (v16i1 x)))
+  // before the setcc result is scalarized on subtargets that don't have legal
+  // vxi1 types.
+  if (DCI.isBeforeLegalize())
+    if (SDValue V = combineBitcastvxi1(DAG, SDValue(N, 0), Subtarget))
+      return V;
   // Since MMX types are special and don't usually play with other vector types,
   // it's better to handle them early to be sure we emit efficient code by
   // avoiding store-load conversions.
 
   // Detect bitcasts between i32 to x86mmx low word.
   if (VT == MVT::x86mmx && N0.getOpcode() == ISD::BUILD_VECTOR &&
       SrcVT == MVT::v2i32 && isNullConstant(N0.getOperand(1))) {
     SDValue N00 = N0->getOperand(0);
SDValue X86TargetLowering::PerformDAGCombine(SDNode *N,
   case X86ISD::PEXTRW:
   case X86ISD::PEXTRB:
     return combineExtractVectorElt_SSE(N, DAG, DCI, Subtarget);
   case ISD::INSERT_SUBVECTOR:
     return combineInsertSubvector(N, DAG, DCI, Subtarget);
   case ISD::VSELECT:
   case ISD::SELECT:
   case X86ISD::SHRUNKBLEND: return combineSelect(N, DAG, DCI, Subtarget);
-  case ISD::BITCAST:        return combineBitcast(N, DAG, Subtarget);
+  case ISD::BITCAST:        return combineBitcast(N, DAG, DCI, Subtarget);
   case X86ISD::CMOV:        return combineCMov(N, DAG, DCI, Subtarget);
   case ISD::ADD:            return combineAdd(N, DAG, Subtarget);
   case ISD::SUB:            return combineSub(N, DAG, Subtarget);
   case X86ISD::ADD:         return combineX86ADD(N, DAG, DCI);
   case X86ISD::ADC:         return combineADC(N, DAG, DCI);
   case ISD::MUL:            return combineMul(N, DAG, DCI, Subtarget);
   case ISD::SHL:
   case ISD::SRA:

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

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+sse2 < %s | FileCheck %s --check-prefixes=SSE2-SSSE3,SSE2
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+ssse3 < %s | FileCheck %s --check-prefixes=SSE2-SSSE3,SSSE3
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx < %s | FileCheck %s --check-prefixes=AVX12,AVX1
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx2 < %s | FileCheck %s --check-prefixes=AVX12,AVX2
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx512f,+avx512vl,+avx512bw < %s | FileCheck %s --check-prefixes=AVX512
 
 define i8 @v8i16(<8 x i16> %a, <8 x i16> %b, <8 x i16> %c, <8 x i16> %d) {
-; SSE2-SSSE3-LABEL: v8i16:
-; SSE2-SSSE3:       ## BB#0:
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    pextrw $7, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $6, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $5, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $4, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $3, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $2, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $1, %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movd %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    retq
+; SSE2-LABEL: v8i16:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE2-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm2
+; SSE2-NEXT:    packuswb %xmm2, %xmm2
+; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v8i16:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSSE3-NEXT:    pand %xmm0, %xmm2
+; SSSE3-NEXT:    pshufb {{.*#+}} xmm2 = xmm2[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSSE3-NEXT:    pmovmskb %xmm2, %eax
+; SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v8i16:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtw %xmm3, %xmm2, %xmm1
 ; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovd %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v8i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtw %xmm1, %xmm0, %k1
 ; AVX512-NEXT:    vpcmpgtw %xmm3, %xmm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <8 x i16> %a, %b
   %x1 = icmp sgt <8 x i16> %c, %d
   %y = and <8 x i1> %x0, %x1
   %res = bitcast <8 x i1> %y to i8
   ret i8 %res
 }
 
 define i4 @v4i32(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c, <4 x i32> %d) {
 ; SSE2-SSSE3-LABEL: v4i32:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    movd %xmm2, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm2, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v4i32:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm1
 ; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovd %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtd %xmm1, %xmm0, %k1
 ; AVX512-NEXT:    vpcmpgtd %xmm3, %xmm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <4 x i32> %a, %b
   %x1 = icmp sgt <4 x i32> %c, %d
   %y = and <4 x i1> %x0, %x1
   %res = bitcast <4 x i1> %y to i4
   ret i4 %res
 }
 
 define i4 @v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c, <4 x float> %d) {
 ; SSE2-SSSE3-LABEL: v4f32:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    cmpltps %xmm0, %xmm1
 ; SSE2-SSSE3-NEXT:    cmpltps %xmm2, %xmm3
 ; SSE2-SSSE3-NEXT:    andps %xmm1, %xmm3
-; SSE2-SSSE3-NEXT:    movd %xmm3, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm3, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v4f32:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vcmpltps %xmm0, %xmm1, %xmm0
 ; AVX12-NEXT:    vcmpltps %xmm2, %xmm3, %xmm1
 ; AVX12-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovd %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4f32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltps %xmm0, %xmm1, %k1
 ; AVX512-NEXT:    vcmpltps %xmm2, %xmm3, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x0 = fcmp ogt <4 x float> %a, %b
   %x1 = fcmp ogt <4 x float> %c, %d
   %y = and <4 x i1> %x0, %x1
   %res = bitcast <4 x i1> %y to i4
   ret i4 %res
 }
 
 define i16 @v16i8(<16 x i8> %a, <16 x i8> %b, <16 x i8> %c, <16 x i8> %d) {
 ; SSE2-SSSE3-LABEL: v16i8:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    movdqa %xmm2, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %cl
-; SSE2-SSSE3-NEXT:    andb $1, %cl
-; SSE2-SSSE3-NEXT:    movb %cl, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v16i8:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtb %xmm3, %xmm2, %xmm1
 ; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrb $15, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $14, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $13, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $12, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $11, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $10, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $9, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $8, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $7, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $6, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $5, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $4, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $3, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $2, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $1, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX12-NEXT:    andb $1, %al
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v16i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtb %xmm1, %xmm0, %k1
 ; AVX512-NEXT:    vpcmpgtb %xmm3, %xmm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>

 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm4, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    pand %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: v2i8:
 ; AVX1:       ## BB#0:
 ; AVX1-NEXT:    vpsllq $56, %xmm3, %xmm3
 ; AVX1-NEXT:    vpsrad $31, %xmm3, %xmm4
 ; AVX1-NEXT:    vpsrad $24, %xmm3, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
 ; AVX1-NEXT:    vpsllq $56, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm3
 ; AVX1-NEXT:    vpsrad $24, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3],xmm0[4,5],xmm3[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX1-NEXT:    vmovmskpd %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: v2i8:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpsllq $56, %xmm3, %xmm3
 ; AVX2-NEXT:    vpsrad $31, %xmm3, %xmm4
 ; AVX2-NEXT:    vpsrad $24, %xmm3, %xmm3
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]

 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2],xmm3[3]
 ; AVX2-NEXT:    vpsllq $56, %xmm0, %xmm0
 ; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm3
 ; AVX2-NEXT:    vpsrad $24, %xmm0, %xmm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2],xmm3[3]
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskpd %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v2i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllq $56, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsraq $56, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsllq $56, %xmm2, %xmm2
 ; AVX512-NEXT:    vpsraq $56, %xmm2, %xmm2

 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm4, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    pand %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: v2i16:
 ; AVX1:       ## BB#0:
 ; AVX1-NEXT:    vpsllq $48, %xmm3, %xmm3
 ; AVX1-NEXT:    vpsrad $31, %xmm3, %xmm4
 ; AVX1-NEXT:    vpsrad $16, %xmm3, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]

 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
 ; AVX1-NEXT:    vpsllq $48, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm3
 ; AVX1-NEXT:    vpsrad $16, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3],xmm0[4,5],xmm3[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX1-NEXT:    vmovmskpd %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: v2i16:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpsllq $48, %xmm3, %xmm3
 ; AVX2-NEXT:    vpsrad $31, %xmm3, %xmm4
 ; AVX2-NEXT:    vpsrad $16, %xmm3, %xmm3
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]

 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2],xmm3[3]
 ; AVX2-NEXT:    vpsllq $48, %xmm0, %xmm0
 ; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm3
 ; AVX2-NEXT:    vpsrad $16, %xmm0, %xmm0
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2],xmm3[3]
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskpd %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v2i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllq $48, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsraq $48, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsllq $48, %xmm2, %xmm2
 ; AVX512-NEXT:    vpsraq $48, %xmm2, %xmm2

 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm2, %xmm4
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm1, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm0
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: v2i32:
 ; AVX1:       ## BB#0:
 ; AVX1-NEXT:    vpsllq $32, %xmm3, %xmm3
 ; AVX1-NEXT:    vpsrad $31, %xmm3, %xmm4
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5],xmm4[6,7]
 ; AVX1-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX1-NEXT:    vpsrad $31, %xmm2, %xmm4
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3],xmm2[4,5],xmm4[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpsllq $32, %xmm1, %xmm1
 ; AVX1-NEXT:    vpsrad $31, %xmm1, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
 ; AVX1-NEXT:    vpsllq $32, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm3
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3],xmm0[4,5],xmm3[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX1-NEXT:    vmovmskpd %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: v2i32:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpsllq $32, %xmm3, %xmm3
 ; AVX2-NEXT:    vpsrad $31, %xmm3, %xmm4
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2],xmm4[3]
 ; AVX2-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX2-NEXT:    vpsrad $31, %xmm2, %xmm4
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0],xmm4[1],xmm2[2],xmm4[3]
 ; AVX2-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
 ; AVX2-NEXT:    vpsllq $32, %xmm1, %xmm1
 ; AVX2-NEXT:    vpsrad $31, %xmm1, %xmm3
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2],xmm3[3]
 ; AVX2-NEXT:    vpsllq $32, %xmm0, %xmm0
 ; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm3
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2],xmm3[3]
 ; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovq %xmm0, %rax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskpd %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v2i32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllq $32, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsraq $32, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsllq $32, %xmm2, %xmm2
 ; AVX512-NEXT:    vpsraq $32, %xmm2, %xmm2

 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm4, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    pand %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v2i64:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm1
 ; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovq %xmm0, %rax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vmovmskpd %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v2i64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtq %xmm1, %xmm0, %k1
 ; AVX512-NEXT:    vpcmpgtq %xmm3, %xmm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <2 x i64> %a, %b
   %x1 = icmp sgt <2 x i64> %c, %d
   %y = and <2 x i1> %x0, %x1
   %res = bitcast <2 x i1> %y to i2
   ret i2 %res
 }
 
 define i2 @v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c, <2 x double> %d) {
 ; SSE2-SSSE3-LABEL: v2f64:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    cmpltpd %xmm0, %xmm1
 ; SSE2-SSSE3-NEXT:    cmpltpd %xmm2, %xmm3
 ; SSE2-SSSE3-NEXT:    andpd %xmm1, %xmm3
-; SSE2-SSSE3-NEXT:    movq %xmm3, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm3, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v2f64:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
 ; AVX12-NEXT:    vcmpltpd %xmm2, %xmm3, %xmm1
 ; AVX12-NEXT:    vandpd %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovq %xmm0, %rax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vmovmskpd %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v2f64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltpd %xmm0, %xmm1, %k1
 ; AVX512-NEXT:    vcmpltpd %xmm2, %xmm3, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)

 ; SSE2-SSSE3-NEXT:    psrad $24, %xmm2
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pslld $24, %xmm1
 ; SSE2-SSSE3-NEXT:    psrad $24, %xmm1
 ; SSE2-SSSE3-NEXT:    pslld $24, %xmm0
 ; SSE2-SSSE3-NEXT:    psrad $24, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pand %xmm2, %xmm0
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v4i8:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpslld $24, %xmm3, %xmm3
 ; AVX12-NEXT:    vpsrad $24, %xmm3, %xmm3
 ; AVX12-NEXT:    vpslld $24, %xmm2, %xmm2
 ; AVX12-NEXT:    vpsrad $24, %xmm2, %xmm2
 ; AVX12-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
 ; AVX12-NEXT:    vpslld $24, %xmm1, %xmm1
 ; AVX12-NEXT:    vpsrad $24, %xmm1, %xmm1
 ; AVX12-NEXT:    vpslld $24, %xmm0, %xmm0
 ; AVX12-NEXT:    vpsrad $24, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovd %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpslld $24, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsrad $24, %xmm3, %xmm3
 ; AVX512-NEXT:    vpslld $24, %xmm2, %xmm2
 ; AVX512-NEXT:    vpsrad $24, %xmm2, %xmm2

 ; SSE2-SSSE3-NEXT:    psrad $16, %xmm2
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm2
 ; SSE2-SSSE3-NEXT:    pslld $16, %xmm1
 ; SSE2-SSSE3-NEXT:    psrad $16, %xmm1
 ; SSE2-SSSE3-NEXT:    pslld $16, %xmm0
 ; SSE2-SSSE3-NEXT:    psrad $16, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pand %xmm2, %xmm0
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v4i16:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpslld $16, %xmm3, %xmm3
 ; AVX12-NEXT:    vpsrad $16, %xmm3, %xmm3
 ; AVX12-NEXT:    vpslld $16, %xmm2, %xmm2
 ; AVX12-NEXT:    vpsrad $16, %xmm2, %xmm2
 ; AVX12-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
 ; AVX12-NEXT:    vpslld $16, %xmm1, %xmm1
 ; AVX12-NEXT:    vpsrad $16, %xmm1, %xmm1
 ; AVX12-NEXT:    vpslld $16, %xmm0, %xmm0
 ; AVX12-NEXT:    vpsrad $16, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovd %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpslld $16, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsrad $16, %xmm3, %xmm3
 ; AVX512-NEXT:    vpslld $16, %xmm2, %xmm2
 ; AVX512-NEXT:    vpsrad $16, %xmm2, %xmm2
; AVX512-NEXT:    retq
   %x0 = icmp sgt <4 x i16> %a, %b
   %x1 = icmp sgt <4 x i16> %c, %d
   %y = and <4 x i1> %x0, %x1
   %res = bitcast <4 x i1> %y to i4
   ret i4 %res
 }
 
 define i8 @v8i8(<8 x i8> %a, <8 x i8> %b, <8 x i8> %c, <8 x i8> %d) {
-; SSE2-SSSE3-LABEL: v8i8:
-; SSE2-SSSE3:       ## BB#0:
-; SSE2-SSSE3-NEXT:    psllw $8, %xmm3
-; SSE2-SSSE3-NEXT:    psraw $8, %xmm3
-; SSE2-SSSE3-NEXT:    psllw $8, %xmm2
-; SSE2-SSSE3-NEXT:    psraw $8, %xmm2
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
-; SSE2-SSSE3-NEXT:    psllw $8, %xmm1
-; SSE2-SSSE3-NEXT:    psraw $8, %xmm1
-; SSE2-SSSE3-NEXT:    psllw $8, %xmm0
-; SSE2-SSSE3-NEXT:    psraw $8, %xmm0
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    pand %xmm2, %xmm0
-; SSE2-SSSE3-NEXT:    pextrw $7, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $6, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $5, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $4, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $3, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $2, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $1, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    retq
+; SSE2-LABEL: v8i8:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    psllw $8, %xmm3
+; SSE2-NEXT:    psraw $8, %xmm3
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    psraw $8, %xmm2
+; SSE2-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm1
+; SSE2-NEXT:    psraw $8, %xmm1
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    psraw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    packuswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v8i8:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    psllw $8, %xmm3
+; SSSE3-NEXT:    psraw $8, %xmm3
+; SSSE3-NEXT:    psllw $8, %xmm2
+; SSSE3-NEXT:    psraw $8, %xmm2
+; SSSE3-NEXT:    pcmpgtw %xmm3, %xmm2
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    psraw $8, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    psraw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSSE3-NEXT:    pmovmskb %xmm0, %eax
+; SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
 ;
 ; AVX12-LABEL: v8i8:
 ; AVX12:       ## BB#0:
 ; AVX12-NEXT:    vpsllw $8, %xmm3, %xmm3
 ; AVX12-NEXT:    vpsraw $8, %xmm3, %xmm3
 ; AVX12-NEXT:    vpsllw $8, %xmm2, %xmm2
 ; AVX12-NEXT:    vpsraw $8, %xmm2, %xmm2
 ; AVX12-NEXT:    vpcmpgtw %xmm3, %xmm2, %xmm2
 ; AVX12-NEXT:    vpsllw $8, %xmm1, %xmm1
 ; AVX12-NEXT:    vpsraw $8, %xmm1, %xmm1
 ; AVX12-NEXT:    vpsllw $8, %xmm0, %xmm0
 ; AVX12-NEXT:    vpsraw $8, %xmm0, %xmm0
 ; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
 ; AVX12-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX12-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    vmovd %xmm0, %eax
-; AVX12-NEXT:    andl $1, %eax
-; AVX12-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX12-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v8i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllw $8, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsraw $8, %xmm3, %xmm3
 ; AVX512-NEXT:    vpsllw $8, %xmm2, %xmm2
 ; AVX512-NEXT:    vpsraw $8, %xmm2, %xmm2

llvm/trunk/test/CodeGen/X86/bitcast-and-setcc-256.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx2 < %s | FileCheck %s --check-prefix=AVX2
+; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+SSE2 < %s | FileCheck %s --check-prefixes=SSE2-SSSE3,SSE2
+; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+SSSE3 < %s | FileCheck %s --check-prefixes=SSE2-SSSE3,SSSE3
+; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx < %s | FileCheck %s --check-prefixes=AVX12,AVX1
+; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx2 < %s | FileCheck %s --check-prefixes=AVX12,AVX2
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx512f,+avx512vl,+avx512bw < %s | FileCheck %s --check-prefix=AVX512
 
 define i4 @v4i64(<4 x i64> %a, <4 x i64> %b, <4 x i64> %c, <4 x i64> %d) {
+; SSE2-SSSE3-LABEL: v4i64:
+; SSE2-SSSE3:       ## BB#0:
+; SSE2-SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [2147483648,0,2147483648,0]
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm3
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm1
+; SSE2-SSSE3-NEXT:    movdqa %xmm1, %xmm9
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm3, %xmm9
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm10 = xmm9[0,0,2,2]
+; SSE2-SSSE3-NEXT:    pcmpeqd %xmm3, %xmm1
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT:    pand %xmm10, %xmm1
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[1,1,3,3]
+; SSE2-SSSE3-NEXT:    por %xmm1, %xmm3
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm2
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm0
+; SSE2-SSSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm2, %xmm1
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[0,0,2,2]
+; SSE2-SSSE3-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE2-SSSE3-NEXT:    pand %xmm9, %xmm2
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT:    por %xmm2, %xmm0
+; SSE2-SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
+; SSE2-SSSE3-NEXT:    pslld $31, %xmm0
+; SSE2-SSSE3-NEXT:    psrad $31, %xmm0
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm7
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm5
+; SSE2-SSSE3-NEXT:    movdqa %xmm5, %xmm1
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm7, %xmm1
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,2,2]
+; SSE2-SSSE3-NEXT:    pcmpeqd %xmm7, %xmm5
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[1,1,3,3]
+; SSE2-SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-SSSE3-NEXT:    por %xmm3, %xmm1
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm6
+; SSE2-SSSE3-NEXT:    pxor %xmm8, %xmm4
+; SSE2-SSSE3-NEXT:    movdqa %xmm4, %xmm2
+; SSE2-SSSE3-NEXT:    pcmpgtd %xmm6, %xmm2
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-SSSE3-NEXT:    pcmpeqd %xmm6, %xmm4
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
+; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm4
+; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE2-SSSE3-NEXT:    por %xmm4, %xmm2
+; SSE2-SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE2-SSSE3-NEXT:    pslld $31, %xmm2
+; SSE2-SSSE3-NEXT:    psrad $31, %xmm2
+; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm2
+; SSE2-SSSE3-NEXT:    movmskps %xmm2, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: v4i64:
+; AVX1:       ## BB#0:
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX1-NEXT:    vpcmpgtq %xmm4, %xmm5, %xmm4
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm4, %xmm1
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovmskps %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
 ; AVX2-LABEL: v4i64:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpgtq %ymm3, %ymm2, %ymm1
 ; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
 ; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskps %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtq %ymm1, %ymm0, %k1
 ; AVX512-NEXT:    vpcmpgtq %ymm3, %ymm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <4 x i64> %a, %b
   %x1 = icmp sgt <4 x i64> %c, %d
   %y = and <4 x i1> %x0, %x1
   %res = bitcast <4 x i1> %y to i4
   ret i4 %res
 }
 
 define i4 @v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x double> %d) {
-; AVX2-LABEL: v4f64:
-; AVX2:       ## BB#0:
-; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vcmpltpd %ymm2, %ymm3, %ymm1
-; AVX2-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; SSE2-SSSE3-LABEL: v4f64:
+; SSE2-SSSE3:       ## BB#0:
+; SSE2-SSSE3-NEXT:    cmpltpd %xmm1, %xmm3
+; SSE2-SSSE3-NEXT:    cmpltpd %xmm0, %xmm2
+; SSE2-SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
+; SSE2-SSSE3-NEXT:    pslld $31, %xmm2
+; SSE2-SSSE3-NEXT:    psrad $31, %xmm2
+; SSE2-SSSE3-NEXT:    cmpltpd %xmm5, %xmm7
+; SSE2-SSSE3-NEXT:    cmpltpd %xmm4, %xmm6
+; SSE2-SSSE3-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[0,2]
+; SSE2-SSSE3-NEXT:    pslld $31, %xmm6
+; SSE2-SSSE3-NEXT:    psrad $31, %xmm6
+; SSE2-SSSE3-NEXT:    pand %xmm2, %xmm6
+; SSE2-SSSE3-NEXT:    movmskps %xmm6, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-SSSE3-NEXT:    retq
+;
+; AVX12-LABEL: v4f64:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
+; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX12-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vcmpltpd %ymm2, %ymm3, %ymm1
+; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX12-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
+; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    vzeroupper
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4f64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltpd %ymm0, %ymm1, %k1
 ; AVX512-NEXT:    vcmpltpd %ymm2, %ymm3, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x0 = fcmp ogt <4 x double> %a, %b
   %x1 = fcmp ogt <4 x double> %c, %d
   %y = and <4 x i1> %x0, %x1
   %res = bitcast <4 x i1> %y to i4
   ret i4 %res
 }
 
 define i16 @v16i16(<16 x i16> %a, <16 x i16> %b, <16 x i16> %c, <16 x i16> %d) {
+; SSE2-LABEL: v16i16:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    pcmpgtw %xmm3, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    pand %xmm3, %xmm0
+; SSE2-NEXT:    packuswb %xmm1, %xmm0
+; SSE2-NEXT:    psllw $7, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm8 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSE2-NEXT:    pand %xmm8, %xmm0
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpgtb %xmm0, %xmm1
+; SSE2-NEXT:    pcmpgtw %xmm7, %xmm5
+; SSE2-NEXT:    pand %xmm3, %xmm5
+; SSE2-NEXT:    pcmpgtw %xmm6, %xmm4
+; SSE2-NEXT:    pand %xmm3, %xmm4
+; SSE2-NEXT:    packuswb %xmm5, %xmm4
+; SSE2-NEXT:    psllw $7, %xmm4
+; SSE2-NEXT:    pand %xmm8, %xmm4
+; SSE2-NEXT:    pcmpgtb %xmm4, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v16i16:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    pcmpgtw %xmm3, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; SSSE3-NEXT:    pshufb %xmm3, %xmm1
+; SSSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm3, %xmm0
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSSE3-NEXT:    psllw $7, %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm8 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSSE3-NEXT:    pand %xmm8, %xmm0
+; SSSE3-NEXT:    pxor %xmm2, %xmm2
+; SSSE3-NEXT:    pxor %xmm1, %xmm1
+; SSSE3-NEXT:    pcmpgtb %xmm0, %xmm1
+; SSSE3-NEXT:    pcmpgtw %xmm7, %xmm5
+; SSSE3-NEXT:    pshufb %xmm3, %xmm5
+; SSSE3-NEXT:    pcmpgtw %xmm6, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm5[0]
+; SSSE3-NEXT:    psllw $7, %xmm4
+; SSSE3-NEXT:    pand %xmm8, %xmm4
+; SSSE3-NEXT:    pcmpgtb %xmm4, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    pmovmskb %xmm2, %eax
+; SSSE3-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: v16i16:
+; AVX1:       ## BB#0:
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX1-NEXT:    vpcmpgtw %xmm4, %xmm5, %xmm4
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm4, %xmm1
+; AVX1-NEXT:    vpcmpgtw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
 ; AVX2-LABEL: v16i16:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpgtw %ymm3, %ymm2, %ymm1
 ; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
 ; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrb $15, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v16i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtw %ymm1, %ymm0, %k1
 ; AVX512-NEXT:    vpcmpgtw %ymm3, %ymm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <16 x i16> %a, %b
   %x1 = icmp sgt <16 x i16> %c, %d
   %y = and <16 x i1> %x0, %x1
   %res = bitcast <16 x i1> %y to i16
   ret i16 %res
 }
 
 define i8 @v8i32(<8 x i32> %a, <8 x i32> %b, <8 x i32> %c, <8 x i32> %d) {
+; SSE2-LABEL: v8i32:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    pcmpgtd %xmm3, %xmm1
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pcmpgtd %xmm2, %xmm0
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE2-NEXT:    psllw $15, %xmm0
+; SSE2-NEXT:    psraw $15, %xmm0
+; SSE2-NEXT:    pcmpgtd %xmm7, %xmm5
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE2-NEXT:    psllw $15, %xmm2
+; SSE2-NEXT:    psraw $15, %xmm2
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm2
+; SSE2-NEXT:    packuswb %xmm2, %xmm2
+; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v8i32:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm3, %xmm1
+; SSSE3-NEXT:    pcmpgtd %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm3, %xmm0
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSSE3-NEXT:    psllw $15, %xmm0
+; SSSE3-NEXT:    psraw $15, %xmm0
+; SSSE3-NEXT:    pcmpgtd %xmm7, %xmm5
+; SSSE3-NEXT:    pshufb %xmm3, %xmm5
+; SSSE3-NEXT:    pcmpgtd %xmm6, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm5[0]
+; SSSE3-NEXT:    psllw $15, %xmm4
+; SSSE3-NEXT:    psraw $15, %xmm4
+; SSSE3-NEXT:    pand %xmm0, %xmm4
+; SSSE3-NEXT:    pshufb {{.*#+}} xmm4 = xmm4[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSSE3-NEXT:    pmovmskb %xmm4, %eax
+; SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: v8i32:
+; AVX1:       ## BB#0:
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX1-NEXT:    vpcmpgtd %xmm4, %xmm5, %xmm4
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm4, %xmm1
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX1-NEXT:    vpmovmskb %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
 ; AVX2-LABEL: v8i32:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm2, %ymm1
 ; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
 ; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v8i32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1
 ; AVX512-NEXT:    vpcmpgtd %ymm3, %ymm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <8 x i32> %a, %b
   %x1 = icmp sgt <8 x i32> %c, %d
   %y = and <8 x i1> %x0, %x1
   %res = bitcast <8 x i1> %y to i8
   ret i8 %res
 }
 
 define i8 @v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x float> %d) {
-; AVX2-LABEL: v8f32:
-; AVX2:       ## BB#0:
-; AVX2-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vcmpltps %ymm2, %ymm3, %ymm1
-; AVX2-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
+; SSE2-LABEL: v8f32:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    cmpltps %xmm1, %xmm3
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    cmpltps %xmm0, %xmm2
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE2-NEXT:    psllw $15, %xmm0
+; SSE2-NEXT:    psraw $15, %xmm0
+; SSE2-NEXT:    cmpltps %xmm5, %xmm7
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE2-NEXT:    cmpltps %xmm4, %xmm6
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[0,2,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE2-NEXT:    psllw $15, %xmm2
+; SSE2-NEXT:    psraw $15, %xmm2
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm2
+; SSE2-NEXT:    packuswb %xmm2, %xmm2
+; SSE2-NEXT:    pmovmskb %xmm2, %eax
+; SSE2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v8f32:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    cmpltps %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSSE3-NEXT:    pshufb %xmm1, %xmm3
+; SSSE3-NEXT:    cmpltps %xmm0, %xmm2
+; SSSE3-NEXT:    pshufb %xmm1, %xmm2
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSSE3-NEXT:    psllw $15, %xmm2
+; SSSE3-NEXT:    psraw $15, %xmm2
+; SSSE3-NEXT:    cmpltps %xmm5, %xmm7
+; SSSE3-NEXT:    pshufb %xmm1, %xmm7
+; SSSE3-NEXT:    cmpltps %xmm4, %xmm6
+; SSSE3-NEXT:    pshufb %xmm1, %xmm6
+; SSSE3-NEXT:    punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm7[0]
+; SSSE3-NEXT:    psllw $15, %xmm6
+; SSSE3-NEXT:    psraw $15, %xmm6
+; SSSE3-NEXT:    pand %xmm2, %xmm6
+; SSSE3-NEXT:    pshufb {{.*#+}} xmm6 = xmm6[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSSE3-NEXT:    pmovmskb %xmm6, %eax
+; SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
+;
+; AVX12-LABEL: v8f32:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
+; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX12-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vcmpltps %ymm2, %ymm3, %ymm1
+; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX12-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
+; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    vzeroupper
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v8f32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltps %ymm0, %ymm1, %k1
 ; AVX512-NEXT:    vcmpltps %ymm2, %ymm3, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x0 = fcmp ogt <8 x float> %a, %b
   %x1 = fcmp ogt <8 x float> %c, %d
   %y = and <8 x i1> %x0, %x1
   %res = bitcast <8 x i1> %y to i8
   ret i8 %res
 }
 
 define i32 @v32i8(<32 x i8> %a, <32 x i8> %b, <32 x i8> %c, <32 x i8> %d) {
+; SSE2-SSSE3-LABEL: v32i8:
+; SSE2-SSSE3:       ## BB#0:
+; SSE2-SSSE3-NEXT:    pcmpgtb %xmm2, %xmm0
+; SSE2-SSSE3-NEXT:    pcmpgtb %xmm3, %xmm1
+; SSE2-SSSE3-NEXT:    pcmpgtb %xmm6, %xmm4
+; SSE2-SSSE3-NEXT:    pand %xmm0, %xmm4
+; SSE2-SSSE3-NEXT:    pcmpgtb %xmm7, %xmm5
+; SSE2-SSSE3-NEXT:    pand %xmm1, %xmm5
+; SSE2-SSSE3-NEXT:    movdqa %xmm5, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movdqa %xmm4, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %cl
+; SSE2-SSSE3-NEXT:    andb $1, %cl
+; SSE2-SSSE3-NEXT:    movb %cl, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %cl
+; SSE2-SSSE3-NEXT:    andb $1, %cl
+; SSE2-SSSE3-NEXT:    movb %cl, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    andb $1, %al
+; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE2-SSSE3-NEXT:    movl -{{[0-9]+}}(%rsp), %ecx
+; SSE2-SSSE3-NEXT:    shll $16, %ecx
+; SSE2-SSSE3-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; SSE2-SSSE3-NEXT:    orl %ecx, %eax
+; SSE2-SSSE3-NEXT:    retq
+;
+; AVX1-LABEL: v32i8:
+; AVX1:       ## BB#0:
+; AVX1-NEXT:    pushq %rbp
+; AVX1-NEXT:  Lcfi0:
+; AVX1-NEXT:    .cfi_def_cfa_offset 16
+; AVX1-NEXT:  Lcfi1:
+; AVX1-NEXT:    .cfi_offset %rbp, -16
+; AVX1-NEXT:    movq %rsp, %rbp
+; AVX1-NEXT:  Lcfi2:
+; AVX1-NEXT:    .cfi_def_cfa_register %rbp
+; AVX1-NEXT:    andq $-32, %rsp
+; AVX1-NEXT:    subq $32, %rsp
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
+; AVX1-NEXT:    vpcmpgtb %xmm4, %xmm5, %xmm4
+; AVX1-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX1-NEXT:    vpcmpgtb %xmm1, %xmm4, %xmm1
+; AVX1-NEXT:    vpcmpgtb %xmm3, %xmm2, %xmm2
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT:    vandps %ymm1, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpextrb $15, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $14, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $13, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $12, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $11, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $10, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $9, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $8, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $7, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $6, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $5, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $4, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $3, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $2, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $1, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $0, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $15, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $14, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $13, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $12, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $11, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $10, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $9, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $8, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $7, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $6, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $5, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $4, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $3, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $2, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $1, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    movl (%rsp), %eax
+; AVX1-NEXT:    movq %rbp, %rsp
+; AVX1-NEXT:    popq %rbp
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
 ; AVX2-LABEL: v32i8:
 ; AVX2:       ## BB#0:
-; AVX2-NEXT:    pushq %rbp
-; AVX2-NEXT:  Lcfi0:
-; AVX2-NEXT:    .cfi_def_cfa_offset 16
-; AVX2-NEXT:  Lcfi1:
-; AVX2-NEXT:    .cfi_offset %rbp, -16
-; AVX2-NEXT:    movq %rsp, %rbp
-; AVX2-NEXT:  Lcfi2:
-; AVX2-NEXT:    .cfi_def_cfa_register %rbp
-; AVX2-NEXT:    andq $-32, %rsp
-; AVX2-NEXT:    subq $32, %rsp
 ; AVX2-NEXT:    vpcmpgtb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpgtb %ymm3, %ymm2, %ymm1
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpextrb $15, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $15, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    movl (%rsp), %eax
-; AVX2-NEXT:    movq %rbp, %rsp
-; AVX2-NEXT:    popq %rbp
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v32i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtb %ymm1, %ymm0, %k1
 ; AVX512-NEXT:    vpcmpgtb %ymm3, %ymm2, %k0 {%k1}
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x0 = icmp sgt <32 x i8> %a, %b
   %x1 = icmp sgt <32 x i8> %c, %d
   %y = and <32 x i1> %x0, %x1
   %res = bitcast <32 x i1> %y to i32
   ret i32 %res
 }

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

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+sse2 < %s | FileCheck %s --check-prefixes=CHECK,SSE2-SSSE3,SSE2
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+ssse3 < %s | FileCheck %s --check-prefixes=CHECK,SSE2-SSSE3,SSSE3
-; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx < %s | FileCheck %s --check-prefixes=CHECK,AVX1
+; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx < %s | FileCheck %s --check-prefixes=CHECK,AVX12,AVX1
+; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx2 < %s | FileCheck %s --check-prefixes=CHECK,AVX12,AVX2
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx512f,+avx512vl,+avx512bw < %s | FileCheck %s --check-prefixes=CHECK,AVX512
 
 define i8 @v8i16(<8 x i16> %a, <8 x i16> %b) {
-; SSE2-SSSE3-LABEL: v8i16:
-; SSE2-SSSE3:       ## BB#0:
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    pextrw $7, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $6, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $5, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $4, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $3, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $2, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $1, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    retq
-;
-; AVX1-LABEL: v8i16:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; SSE2-LABEL: v8i16:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    packuswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v8i16:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSSE3-NEXT:    pmovmskb %xmm0, %eax
+; SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
+;
+; AVX12-LABEL: v8i16:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v8i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    retq
   %x = icmp sgt <8 x i16> %a, %b
   %res = bitcast <8 x i1> %x to i8
   ret i8 %res
 }
 
 define i4 @v4i32(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-SSSE3-LABEL: v4i32:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v4i32:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v4i32:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtd %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x = icmp sgt <4 x i32> %a, %b
   %res = bitcast <4 x i1> %x to i4
   ret i4 %res
 }
 
 define i4 @v4f32(<4 x float> %a, <4 x float> %b) {
 ; SSE2-SSSE3-LABEL: v4f32:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    cmpltps %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movaps %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm1, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v4f32:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vcmpltps %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vextractps $3, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vextractps $2, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vextractps $1, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vextractps $0, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v4f32:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vcmpltps %xmm0, %xmm1, %xmm0
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4f32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltps %xmm0, %xmm1, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x = fcmp ogt <4 x float> %a, %b
   %res = bitcast <4 x i1> %x to i4
   ret i4 %res
 }
 
 define i16 @v16i8(<16 x i8> %a, <16 x i8> %b) {
 ; SSE2-SSSE3-LABEL: v16i8:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    pcmpgtb %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movdqa %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %cl
-; SSE2-SSSE3-NEXT:    andb $1, %cl
-; SSE2-SSSE3-NEXT:    movb %cl, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    andb $1, %al
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; SSE2-SSSE3-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v16i8:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrb $15, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $14, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $13, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $12, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $11, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $10, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $9, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $8, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $7, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $6, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $5, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $4, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $3, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $2, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $1, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v16i8:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v16i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX512-NEXT:    retq
   %x = icmp sgt <16 x i8> %a, %b

 ; SSE2-SSSE3-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    movq %xmm1, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm1, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: v2i8:
 ; AVX1:       ## BB#0:
 ; AVX1-NEXT:    vpsllq $56, %xmm1, %xmm1
 ; AVX1-NEXT:    vpsrad $31, %xmm1, %xmm2
 ; AVX1-NEXT:    vpsrad $24, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpsllq $56, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm2
 ; AVX1-NEXT:    vpsrad $24, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX1-NEXT:    vmovmskpd %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX1-NEXT:    retq
 ;
+; AVX2-LABEL: v2i8:
+; AVX2:       ## BB#0:
+; AVX2-NEXT:    vpsllq $56, %xmm1, %xmm1
+; AVX2-NEXT:    vpsrad $31, %xmm1, %xmm2
+; AVX2-NEXT:    vpsrad $24, %xmm1, %xmm1
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; AVX2-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm2
+; AVX2-NEXT:    vpsrad $24, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
+; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovmskpd %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX2-NEXT:    retq
+;
 ; AVX512-LABEL: v2i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllq $56, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsraq $56, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsllq $56, %xmm0, %xmm0
 ; AVX512-NEXT:    vpsraq $56, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax

 ; SSE2-SSSE3-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    movq %xmm1, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm1, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: v2i16:
 ; AVX1:       ## BB#0:
 ; AVX1-NEXT:    vpsllq $48, %xmm1, %xmm1
 ; AVX1-NEXT:    vpsrad $31, %xmm1, %xmm2
 ; AVX1-NEXT:    vpsrad $16, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpsllq $48, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm2
 ; AVX1-NEXT:    vpsrad $16, %xmm0, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX1-NEXT:    vmovmskpd %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX1-NEXT:    retq
 ;
+; AVX2-LABEL: v2i16:
+; AVX2:       ## BB#0:
+; AVX2-NEXT:    vpsllq $48, %xmm1, %xmm1
+; AVX2-NEXT:    vpsrad $31, %xmm1, %xmm2
+; AVX2-NEXT:    vpsrad $16, %xmm1, %xmm1
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; AVX2-NEXT:    vpsllq $48, %xmm0, %xmm0
+; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm2
+; AVX2-NEXT:    vpsrad $16, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
+; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovmskpd %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX2-NEXT:    retq
+;
 ; AVX512-LABEL: v2i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllq $48, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsraq $48, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsllq $48, %xmm0, %xmm0
 ; AVX512-NEXT:    vpsraq $48, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax

 ; SSE2-SSSE3-NEXT:    movdqa %xmm2, %xmm1
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm0, %xmm1
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm0, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    movq %xmm1, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm1, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: v2i32:
 ; AVX1:       ## BB#0:
 ; AVX1-NEXT:    vpsllq $32, %xmm1, %xmm1
 ; AVX1-NEXT:    vpsrad $31, %xmm1, %xmm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpsllq $32, %xmm0, %xmm0
 ; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm2
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
 ; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX1-NEXT:    vmovmskpd %xmm0, %eax
+; AVX1-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX1-NEXT:    retq
 ;
+; AVX2-LABEL: v2i32:
+; AVX2:       ## BB#0:
+; AVX2-NEXT:    vpsllq $32, %xmm1, %xmm1
+; AVX2-NEXT:    vpsrad $31, %xmm1, %xmm2
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; AVX2-NEXT:    vpsllq $32, %xmm0, %xmm0
+; AVX2-NEXT:    vpsrad $31, %xmm0, %xmm2
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX2-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
+; AVX2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovmskpd %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX2-NEXT:    retq
+;
 ; AVX512-LABEL: v2i32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllq $32, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsraq $32, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsllq $32, %xmm0, %xmm0
 ; AVX512-NEXT:    vpsraq $32, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax

 ; SSE2-SSSE3-NEXT:    movdqa %xmm0, %xmm2
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
 ; SSE2-SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; SSE2-SSSE3-NEXT:    por %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    movq %xmm1, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm1, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v2i64:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v2i64:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vmovmskpd %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v2i64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x = icmp sgt <2 x i64> %a, %b
   %res = bitcast <2 x i1> %x to i2
   ret i2 %res
 }
 
 define i2 @v2f64(<2 x double> %a, <2 x double> %b) {
 ; SSE2-SSSE3-LABEL: v2f64:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    cmpltpd %xmm0, %xmm1
-; SSE2-SSSE3-NEXT:    movq %xmm1, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movq %xmm0, %rax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskpd %xmm1, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v2f64:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovq %xmm0, %rax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v2f64:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
+; AVX12-NEXT:    vmovmskpd %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v2f64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltpd %xmm0, %xmm1, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x = fcmp ogt <2 x double> %a, %b
   %res = bitcast <2 x i1> %x to i2
   ret i2 %res
 }
 
 define i4 @v4i8(<4 x i8> %a, <4 x i8> %b) {
 ; SSE2-SSSE3-LABEL: v4i8:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    pslld $24, %xmm1
 ; SSE2-SSSE3-NEXT:    psrad $24, %xmm1
 ; SSE2-SSSE3-NEXT:    pslld $24, %xmm0
 ; SSE2-SSSE3-NEXT:    psrad $24, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v4i8:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpslld $24, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrad $24, %xmm1, %xmm1
-; AVX1-NEXT:    vpslld $24, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrad $24, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v4i8:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpslld $24, %xmm1, %xmm1
+; AVX12-NEXT:    vpsrad $24, %xmm1, %xmm1
+; AVX12-NEXT:    vpslld $24, %xmm0, %xmm0
+; AVX12-NEXT:    vpsrad $24, %xmm0, %xmm0
+; AVX12-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpslld $24, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsrad $24, %xmm1, %xmm1
 ; AVX512-NEXT:    vpslld $24, %xmm0, %xmm0
 ; AVX512-NEXT:    vpsrad $24, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtd %xmm1, %xmm0, %k0

 define i4 @v4i16(<4 x i16> %a, <4 x i16> %b) {
 ; SSE2-SSSE3-LABEL: v4i16:
 ; SSE2-SSSE3:       ## BB#0:
 ; SSE2-SSSE3-NEXT:    pslld $16, %xmm1
 ; SSE2-SSSE3-NEXT:    psrad $16, %xmm1
 ; SSE2-SSSE3-NEXT:    pslld $16, %xmm0
 ; SSE2-SSSE3-NEXT:    psrad $16, %xmm0
 ; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-SSSE3-NEXT:    movd %xmm1, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-SSSE3-NEXT:    movmskps %xmm0, %eax
+; SSE2-SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-SSSE3-NEXT:    retq
 ;
-; AVX1-LABEL: v4i16:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpslld $16, %xmm1, %xmm1
-; AVX1-NEXT:    vpsrad $16, %xmm1, %xmm1
-; AVX1-NEXT:    vpslld $16, %xmm0, %xmm0
-; AVX1-NEXT:    vpsrad $16, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; AVX12-LABEL: v4i16:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpslld $16, %xmm1, %xmm1
+; AVX12-NEXT:    vpsrad $16, %xmm1, %xmm1
+; AVX12-NEXT:    vpslld $16, %xmm0, %xmm0
+; AVX12-NEXT:    vpsrad $16, %xmm0, %xmm0
+; AVX12-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vmovmskps %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpslld $16, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsrad $16, %xmm1, %xmm1
 ; AVX512-NEXT:    vpslld $16, %xmm0, %xmm0
 ; AVX512-NEXT:    vpsrad $16, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtd %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    retq
   %x = icmp sgt <4 x i16> %a, %b
   %res = bitcast <4 x i1> %x to i4
   ret i4 %res
 }
 
 define i8 @v8i8(<8 x i8> %a, <8 x i8> %b) {
-; SSE2-SSSE3-LABEL: v8i8:
-; SSE2-SSSE3:       ## BB#0:
-; SSE2-SSSE3-NEXT:    psllw $8, %xmm1
-; SSE2-SSSE3-NEXT:    psraw $8, %xmm1
-; SSE2-SSSE3-NEXT:    psllw $8, %xmm0
-; SSE2-SSSE3-NEXT:    psraw $8, %xmm0
-; SSE2-SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
-; SSE2-SSSE3-NEXT:    pextrw $7, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $6, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $5, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $4, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $3, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $2, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    pextrw $1, %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
-; SSE2-SSSE3-NEXT:    andl $1, %eax
-; SSE2-SSSE3-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE2-SSSE3-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-SSSE3-NEXT:    retq
-;
-; AVX1-LABEL: v8i8:
-; AVX1:       ## BB#0:
-; AVX1-NEXT:    vpsllw $8, %xmm1, %xmm1
-; AVX1-NEXT:    vpsraw $8, %xmm1, %xmm1
-; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpsraw $8, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vmovd %xmm0, %eax
-; AVX1-NEXT:    andl $1, %eax
-; AVX1-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX1-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; AVX1-NEXT:    retq
+; SSE2-LABEL: v8i8:
+; SSE2:       ## BB#0:
+; SSE2-NEXT:    psllw $8, %xmm1
+; SSE2-NEXT:    psraw $8, %xmm1
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    psraw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    packuswb %xmm0, %xmm0
+; SSE2-NEXT:    pmovmskb %xmm0, %eax
+; SSE2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: v8i8:
+; SSSE3:       ## BB#0:
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    psraw $8, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    psraw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSSE3-NEXT:    pmovmskb %xmm0, %eax
+; SSSE3-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; SSSE3-NEXT:    retq
+;
+; AVX12-LABEL: v8i8:
+; AVX12:       ## BB#0:
+; AVX12-NEXT:    vpsllw $8, %xmm1, %xmm1
+; AVX12-NEXT:    vpsraw $8, %xmm1, %xmm1
+; AVX12-NEXT:    vpsllw $8, %xmm0, %xmm0
+; AVX12-NEXT:    vpsraw $8, %xmm0, %xmm0
+; AVX12-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX12-NEXT:    retq
 ;
 ; AVX512-LABEL: v8i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpsllw $8, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsraw $8, %xmm1, %xmm1
 ; AVX512-NEXT:    vpsllw $8, %xmm0, %xmm0
 ; AVX512-NEXT:    vpsraw $8, %xmm0, %xmm0
 ; AVX512-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    retq
   %x = icmp sgt <8 x i8> %a, %b
   %res = bitcast <8 x i1> %x to i8
   ret i8 %res
 }

llvm/trunk/test/CodeGen/X86/bitcast-setcc-256.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx2 < %s | FileCheck %s --check-prefix=AVX2
 ; RUN: llc -mtriple=x86_64-apple-darwin -mcpu=x86-64 -mattr=+avx512f,+avx512vl,+avx512bw < %s | FileCheck %s --check-prefix=AVX512
 
 define i16 @v16i16(<16 x i16> %a, <16 x i16> %b) {
 ; AVX2-LABEL: v16i16:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrb $15, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v16i16:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtw %ymm1, %ymm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x = icmp sgt <16 x i16> %a, %b
   %res = bitcast <16 x i1> %x to i16
   ret i16 %res
 }
 
 define i8 @v8i32(<8 x i32> %a, <8 x i32> %b) {
 ; AVX2-LABEL: v8i32:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v8i32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x = icmp sgt <8 x i32> %a, %b
   %res = bitcast <8 x i1> %x to i8
   ret i8 %res
 }
 
 define i8 @v8f32(<8 x float> %a, <8 x float> %b) {
 ; AVX2-LABEL: v8f32:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vcmpltps %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v8f32:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltps %ymm0, %ymm1, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x = fcmp ogt <8 x float> %a, %b
   %res = bitcast <8 x i1> %x to i8
   ret i8 %res
 }
 
 define i32 @v32i8(<32 x i8> %a, <32 x i8> %b) {
 ; AVX2-LABEL: v32i8:
 ; AVX2:       ## BB#0:
-; AVX2-NEXT:    pushq %rbp
-; AVX2-NEXT:  Lcfi0:
-; AVX2-NEXT:    .cfi_def_cfa_offset 16
-; AVX2-NEXT:  Lcfi1:
-; AVX2-NEXT:    .cfi_offset %rbp, -16
-; AVX2-NEXT:    movq %rsp, %rbp
-; AVX2-NEXT:  Lcfi2:
-; AVX2-NEXT:    .cfi_def_cfa_register %rbp
-; AVX2-NEXT:    andq $-32, %rsp
-; AVX2-NEXT:    subq $32, %rsp
 ; AVX2-NEXT:    vpcmpgtb %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpextrb $15, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $15, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    movl (%rsp), %eax
-; AVX2-NEXT:    movq %rbp, %rsp
-; AVX2-NEXT:    popq %rbp
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v32i8:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x = icmp sgt <32 x i8> %a, %b
   %res = bitcast <32 x i1> %x to i32
   ret i32 %res
 }
 
 define i4 @v4i64(<4 x i64> %a, <4 x i64> %b) {
 ; AVX2-LABEL: v4i64:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskpd %ymm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v4i64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpcmpgtq %ymm1, %ymm0, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x = icmp sgt <4 x i64> %a, %b
   %res = bitcast <4 x i1> %x to i4
   ret i4 %res
 }
 
 define i4 @v4f64(<4 x double> %a, <4 x double> %b) {
 ; AVX2-LABEL: v4f64:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    vcmpltpd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpextrd $3, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $2, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrd $1, %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vmovd %xmm0, %eax
-; AVX2-NEXT:    andl $1, %eax
-; AVX2-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; AVX2-NEXT:    vmovmskpd %ymm0, %eax
+; AVX2-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: v4f64:
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vcmpltpd %ymm0, %ymm1, %k0
 ; AVX512-NEXT:    kmovd %k0, %eax
 ; AVX512-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
 ; AVX512-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %x = fcmp ogt <4 x double> %a, %b
   %res = bitcast <4 x i1> %x to i4
   ret i4 %res
 }

llvm/trunk/test/CodeGen/X86/setcc-lowering.ll

entry:
   %3 = sext <8 x i1> %2 to <8 x i16>
   ret <8 x i16> %3
 }
 
 define void @pr26232(i64 %a, <16 x i1> %b) {
 ; AVX-LABEL: pr26232:
 ; AVX:       # BB#0: # %for_loop599.preheader
 ; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX-NEXT:    vmovdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
 ; AVX-NEXT:    .p2align 4, 0x90
 ; AVX-NEXT:  .LBB1_1: # %for_loop599
 ; AVX-NEXT:    # =>This Inner Loop Header: Depth=1
 ; AVX-NEXT:    xorl %eax, %eax
 ; AVX-NEXT:    cmpq $65536, %rdi # imm = 0x10000
 ; AVX-NEXT:    setl %al
-; AVX-NEXT:    vmovd %eax, %xmm2
-; AVX-NEXT:    vpshufb %xmm1, %xmm2, %xmm2
-; AVX-NEXT:    vpand %xmm0, %xmm2, %xmm2
-; AVX-NEXT:    vpextrb $15, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $14, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $13, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $12, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $11, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $10, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $9, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $8, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $7, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $6, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $5, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $4, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $3, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $2, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $1, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    vpextrb $0, %xmm2, %eax
-; AVX-NEXT:    andb $1, %al
-; AVX-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; AVX-NEXT:    cmpw $0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    vmovd %eax, %xmm3
+; AVX-NEXT:    vpshufb %xmm1, %xmm3, %xmm3
+; AVX-NEXT:    vpand %xmm0, %xmm3, %xmm3
+; AVX-NEXT:    vpsllw $7, %xmm3, %xmm3
+; AVX-NEXT:    vpand %xmm2, %xmm3, %xmm3
+; AVX-NEXT:    vpcmpgtb %xmm3, %xmm1, %xmm3
+; AVX-NEXT:    vpmovmskb %xmm3, %eax
+; AVX-NEXT:    testw %ax, %ax
 ; AVX-NEXT:    jne .LBB1_1
 ; AVX-NEXT:  # BB#2: # %for_exit600
 ; AVX-NEXT:    retq
 ;
 ; KNL-32-LABEL: pr26232:
 ; KNL-32:       # BB#0: # %for_loop599.preheader
 ; KNL-32-NEXT:    pushl %esi
 ; KNL-32-NEXT:  .Lcfi0: