Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -34261,7 +34261,7 @@
   EVT ExtractVT = Extract->getValueType(0);
   unsigned BitWidth = ExtractVT.getSizeInBits();
   if (ExtractVT != MVT::i64 && ExtractVT != MVT::i32 && ExtractVT != MVT::i16 &&
-      ExtractVT != MVT::i8)
+      ExtractVT != MVT::i8 && ExtractVT != MVT::i1)
     return SDValue();
 
   // Check for OR(any_of) and AND(all_of) horizontal reduction patterns.
@@ -34275,36 +34275,63 @@
   if (Match.getScalarValueSizeInBits() != BitWidth)
     return SDValue();
 
-  unsigned MatchSizeInBits = Match.getValueSizeInBits();
-  if (!(MatchSizeInBits == 128 || (MatchSizeInBits == 256 && Subtarget.hasAVX())))
-    return SDValue();
+  SDValue Movmsk;
+  SDLoc DL(Extract);
+  unsigned NumElts = Match.getValueType().getVectorNumElements();
 
-  // Make sure this isn't a vector of 1 element. The perf win from using MOVMSK
-  // diminishes with less elements in the reduction, but it is generally better
-  // to get the comparison over to the GPRs as soon as possible to reduce the
-  // number of vector ops.
-  if (Match.getValueType().getVectorNumElements() < 2)
-    return SDValue();
+  if (ExtractVT == MVT::i1) {
+    // Special case for (pre-legalization) vXi1 reductions.
+    // Use combineBitcastvxi1 to create the MOVMSK.
+    if (NumElts > 32)
+      return SDValue();
+    if (NumElts == 32 && !Subtarget.hasInt256()) {
+      SDValue Lo, Hi;
+      std::tie(Lo, Hi) = DAG.SplitVector(Match, DL);
+      Match = DAG.getNode(BinOp, DL, Lo.getValueType(), Lo, Hi);
+      NumElts = 16;
+    }
+    EVT MovmskVT = EVT::getIntegerVT(*DAG.getContext(), NumElts);
+    Movmsk = combineBitcastvxi1(DAG, MovmskVT, Match, DL, Subtarget);
+    if (!Movmsk)
+      return SDValue();
+    Movmsk = DAG.getZExtOrTrunc(Movmsk, DL, MVT::i32);
+  } else {
+    unsigned MatchSizeInBits = Match.getValueSizeInBits();
+    if (!(MatchSizeInBits == 128 ||
+          (MatchSizeInBits == 256 && Subtarget.hasAVX())))
+      return SDValue();
 
-  // Check that we are extracting a reduction of all sign bits.
-  if (DAG.ComputeNumSignBits(Match) != BitWidth)
-    return SDValue();
+    // Make sure this isn't a vector of 1 element. The perf win from using
+    // MOVMSK diminishes with less elements in the reduction, but it is
+    // generally better to get the comparison over to the GPRs as soon as
+    // possible to reduce the number of vector ops.
+    if (Match.getValueType().getVectorNumElements() < 2)
+      return SDValue();
 
-  SDLoc DL(Extract);
-  if (MatchSizeInBits == 256 && BitWidth < 32 && !Subtarget.hasInt256()) {
-    SDValue Lo, Hi;
-    std::tie(Lo, Hi) = DAG.SplitVector(Match, DL);
-    Match = DAG.getNode(BinOp, DL, Lo.getValueType(), Lo, Hi);
-    MatchSizeInBits = Match.getValueSizeInBits();
-  }
+    // Check that we are extracting a reduction of all sign bits.
+    if (DAG.ComputeNumSignBits(Match) != BitWidth)
+      return SDValue();
 
-  // For 32/64 bit comparisons use MOVMSKPS/MOVMSKPD, else PMOVMSKB.
-  MVT MaskSrcVT;
-  if (64 == BitWidth || 32 == BitWidth)
-    MaskSrcVT = MVT::getVectorVT(MVT::getFloatingPointVT(BitWidth),
-                                 MatchSizeInBits / BitWidth);
-  else
-    MaskSrcVT = MVT::getVectorVT(MVT::i8, MatchSizeInBits / 8);
+    if (MatchSizeInBits == 256 && BitWidth < 32 && !Subtarget.hasInt256()) {
+      SDValue Lo, Hi;
+      std::tie(Lo, Hi) = DAG.SplitVector(Match, DL);
+      Match = DAG.getNode(BinOp, DL, Lo.getValueType(), Lo, Hi);
+      MatchSizeInBits = Match.getValueSizeInBits();
+    }
+
+    // For 32/64 bit comparisons use MOVMSKPS/MOVMSKPD, else PMOVMSKB.
+    MVT MaskSrcVT;
+    if (64 == BitWidth || 32 == BitWidth)
+      MaskSrcVT = MVT::getVectorVT(MVT::getFloatingPointVT(BitWidth),
+                                   MatchSizeInBits / BitWidth);
+    else
+      MaskSrcVT = MVT::getVectorVT(MVT::i8, MatchSizeInBits / 8);
+
+    SDValue BitcastLogicOp = DAG.getBitcast(MaskSrcVT, Match);
+    Movmsk = getPMOVMSKB(DL, BitcastLogicOp, DAG, Subtarget);
+    NumElts = MaskSrcVT.getVectorNumElements();
+  }
+  assert(NumElts <= 32 && "Not expecting more than 32 elements");
 
   SDValue CmpC;
   ISD::CondCode CondCode;
@@ -34314,8 +34341,6 @@
     CondCode = ISD::CondCode::SETNE;
   } else {
     // all_of -> MOVMSK == ((1 << NumElts) - 1)
-    uint64_t NumElts = MaskSrcVT.getVectorNumElements();
-    assert(NumElts <= 32 && "Not expecting more than 32 elements");
     CmpC = DAG.getConstant((1ULL << NumElts) - 1, DL, MVT::i32);
     CondCode = ISD::CondCode::SETEQ;
   }
@@ -34323,10 +34348,8 @@
   // The setcc produces an i8 of 0/1, so extend that to the result width and
   // negate to get the final 0/-1 mask value.
   const TargetLowering &TLI = DAG.getTargetLoweringInfo();
-  SDValue BitcastLogicOp = DAG.getBitcast(MaskSrcVT, Match);
-  SDValue Movmsk = getPMOVMSKB(DL, BitcastLogicOp, DAG, Subtarget);
-  EVT SetccVT = TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(),
-                                       MVT::i32);
+  EVT SetccVT =
+      TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), MVT::i32);
   SDValue Setcc = DAG.getSetCC(DL, SetccVT, Movmsk, CmpC, CondCode);
   SDValue Zext = DAG.getZExtOrTrunc(Setcc, DL, ExtractVT);
   SDValue Zero = DAG.getConstant(0, DL, ExtractVT);
Index: test/CodeGen/X86/vector-compare-all_of.ll
===================================================================
--- test/CodeGen/X86/vector-compare-all_of.ll
+++ test/CodeGen/X86/vector-compare-all_of.ll
@@ -920,19 +920,17 @@
 ; SSE-LABEL: bool_reduction_v2f64:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskpd %xmm1, %eax
+; SSE-NEXT:    cmpb $3, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v2f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vandpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    cmpb $3, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v2f64:
@@ -957,23 +955,17 @@
 ; SSE-LABEL: bool_reduction_v4f32:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    cmpeqps %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    cmpb $15, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v4f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmpeqps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vandpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vandpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    cmpb $15, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v4f32:
@@ -1005,27 +997,18 @@
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    cmplepd %xmm1, %xmm3
 ; SSE-NEXT:    cmplepd %xmm0, %xmm2
-; SSE-NEXT:    movapd %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm2[0,2]
-; SSE-NEXT:    andps %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packssdw %xmm3, %xmm2
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    cmpb $15, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v4f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmplepd %ymm0, %ymm1, %ymm0
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2
-; AVX-NEXT:    vpackssdw %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskpd %ymm0, %eax
+; AVX-NEXT:    cmpb $15, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
 ;
@@ -1119,10 +1102,9 @@
 ; SSE-NEXT:    pxor %xmm2, %xmm1
 ; SSE-NEXT:    pxor %xmm2, %xmm0
 ; SSE-NEXT:    pcmpgtq %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pextrb $0, %xmm1, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskpd %xmm0, %eax
+; SSE-NEXT:    cmpb $3, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v2i64:
@@ -1131,10 +1113,9 @@
 ; AVX-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpxor %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    cmpb $3, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v2i64:
@@ -1161,25 +1142,19 @@
 ; SSE-NEXT:    pcmpeqd %xmm1, %xmm0
 ; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
 ; SSE-NEXT:    pxor %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pextrb $0, %xmm1, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskps %xmm1, %eax
+; SSE-NEXT:    cmpb $15, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v4i32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpxor %xmm1, %xmm0, %xmm1
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; AVX-NEXT:    vpandn %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    cmpb $15, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v4i32:
@@ -1210,28 +1185,19 @@
 ; SSE-LABEL: bool_reduction_v8i16:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packsswb %xmm0, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
+; SSE-NEXT:    cmpb $-1, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v8i16:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    cmpb $-1, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v8i16:
@@ -1270,18 +1236,16 @@
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pcmpgtb %xmm1, %xmm0
 ; SSE-NEXT:    pmovmskb %xmm0, %eax
-; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    cmpw $-1, %ax
 ; SSE-NEXT:    sete %al
-; SSE-NEXT:    negb %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v16i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpmovmskb %xmm0, %eax
-; AVX-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX-NEXT:    cmpw $-1, %ax
 ; AVX-NEXT:    sete %al
-; AVX-NEXT:    negb %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v16i8:
@@ -1316,13 +1280,10 @@
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pcmpgtq %xmm1, %xmm3
 ; SSE-NEXT:    pcmpgtq %xmm0, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pextrb $0, %xmm1, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packssdw %xmm3, %xmm2
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    cmpb $15, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v4i64:
@@ -1331,27 +1292,19 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm1
-; AVX1-NEXT:    vpackssdw %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX1-NEXT:    # kill: def $al killed $al killed $eax
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmovmskpd %ymm0, %eax
+; AVX1-NEXT:    cmpb $15, %al
+; AVX1-NEXT:    sete %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: bool_reduction_v4i64:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT:    vpackssdw %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    # kill: def $al killed $al killed $eax
+; AVX2-NEXT:    vmovmskpd %ymm0, %eax
+; AVX2-NEXT:    cmpb $15, %al
+; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1388,15 +1341,10 @@
 ; SSE-NEXT:    pminud %xmm0, %xmm2
 ; SSE-NEXT:    pcmpeqd %xmm0, %xmm2
 ; SSE-NEXT:    packssdw %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packsswb %xmm0, %xmm2
+; SSE-NEXT:    pmovmskb %xmm2, %eax
+; SSE-NEXT:    cmpb $-1, %al
+; SSE-NEXT:    sete %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v8i32:
@@ -1407,15 +1355,10 @@
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX1-NEXT:    # kill: def $al killed $al killed $eax
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmovmskps %ymm0, %eax
+; AVX1-NEXT:    cmpb $-1, %al
+; AVX1-NEXT:    sete %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -1423,16 +1366,9 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpminud %ymm1, %ymm0, %ymm1
 ; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    # kill: def $al killed $al killed $eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    cmpb $-1, %al
+; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1474,9 +1410,8 @@
 ; SSE-NEXT:    pcmpeqw %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
 ; SSE-NEXT:    pmovmskb %xmm0, %eax
-; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    cmpw $-1, %ax
 ; SSE-NEXT:    sete %al
-; SSE-NEXT:    negb %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v16i16:
@@ -1487,9 +1422,8 @@
 ; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
-; AVX1-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX1-NEXT:    cmpw $-1, %ax
 ; AVX1-NEXT:    sete %al
-; AVX1-NEXT:    negb %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -1499,9 +1433,8 @@
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
-; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX2-NEXT:    cmpw $-1, %ax
 ; AVX2-NEXT:    sete %al
-; AVX2-NEXT:    negb %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1536,13 +1469,12 @@
 define i1 @bool_reduction_v32i8(<32 x i8> %x, <32 x i8> %y) {
 ; SSE-LABEL: bool_reduction_v32i8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqb %xmm2, %xmm0
 ; SSE-NEXT:    pcmpeqb %xmm3, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pmovmskb %xmm1, %eax
-; SSE-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; SSE-NEXT:    pcmpeqb %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    cmpw $-1, %ax
 ; SSE-NEXT:    sete %al
-; SSE-NEXT:    negb %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v32i8:
@@ -1553,9 +1485,8 @@
 ; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
-; AVX1-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX1-NEXT:    cmpw $-1, %ax
 ; AVX1-NEXT:    sete %al
-; AVX1-NEXT:    negb %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -1565,7 +1496,6 @@
 ; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    cmpl $-1, %eax
 ; AVX2-NEXT:    sete %al
-; AVX2-NEXT:    negb %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
Index: test/CodeGen/X86/vector-compare-any_of.ll
===================================================================
--- test/CodeGen/X86/vector-compare-any_of.ll
+++ test/CodeGen/X86/vector-compare-any_of.ll
@@ -840,19 +840,17 @@
 ; SSE-LABEL: bool_reduction_v2f64:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    cmpltpd %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskpd %xmm1, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v2f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vorpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    testb %al, %al
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v2f64:
@@ -878,23 +876,17 @@
 ; SSE-LABEL: bool_reduction_v4f32:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    cmpeqps %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskps %xmm0, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v4f32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmpeqps %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT:    vorpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vorpd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    testb %al, %al
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v4f32:
@@ -928,27 +920,18 @@
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    cmplepd %xmm1, %xmm3
 ; SSE-NEXT:    cmplepd %xmm0, %xmm2
-; SSE-NEXT:    movapd %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm2[0,2]
-; SSE-NEXT:    orps %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,2,3]
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packssdw %xmm3, %xmm2
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v4f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmplepd %ymm0, %ymm1, %ymm0
-; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2
-; AVX-NEXT:    vpackssdw %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpor %xmm2, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskpd %ymm0, %eax
+; AVX-NEXT:    testb %al, %al
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq
 ;
@@ -1046,10 +1029,9 @@
 ; SSE-NEXT:    pxor %xmm2, %xmm1
 ; SSE-NEXT:    pxor %xmm2, %xmm0
 ; SSE-NEXT:    pcmpgtq %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pextrb $0, %xmm1, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskpd %xmm0, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v2i64:
@@ -1058,10 +1040,9 @@
 ; AVX-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; AVX-NEXT:    vpxor %xmm2, %xmm0, %xmm0
 ; AVX-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    testb %al, %al
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v2i64:
@@ -1089,12 +1070,9 @@
 ; SSE-NEXT:    pcmpeqd %xmm1, %xmm0
 ; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
 ; SSE-NEXT:    pxor %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pextrb $0, %xmm1, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    movmskps %xmm1, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v4i32:
@@ -1102,12 +1080,9 @@
 ; AVX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vmovmskps %xmm0, %eax
+; AVX-NEXT:    testb %al, %al
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v4i32:
@@ -1140,28 +1115,19 @@
 ; SSE-LABEL: bool_reduction_v8i16:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pcmpgtw %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packsswb %xmm0, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v8i16:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    # kill: def $al killed $al killed $eax
+; AVX-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    vpmovmskb %xmm0, %eax
+; AVX-NEXT:    testb %al, %al
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v8i16:
@@ -1203,16 +1169,16 @@
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pcmpgtb %xmm1, %xmm0
 ; SSE-NEXT:    pmovmskb %xmm0, %eax
-; SSE-NEXT:    negl %eax
-; SSE-NEXT:    sbbb %al, %al
+; SSE-NEXT:    testw %ax, %ax
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: bool_reduction_v16i8:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpmovmskb %xmm0, %eax
-; AVX-NEXT:    negl %eax
-; AVX-NEXT:    sbbb %al, %al
+; AVX-NEXT:    testw %ax, %ax
+; AVX-NEXT:    setne %al
 ; AVX-NEXT:    retq
 ;
 ; AVX512-LABEL: bool_reduction_v16i8:
@@ -1247,13 +1213,10 @@
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pcmpgtq %xmm1, %xmm3
 ; SSE-NEXT:    pcmpgtq %xmm0, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pextrb $0, %xmm1, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packssdw %xmm3, %xmm2
+; SSE-NEXT:    movmskps %xmm2, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v4i64:
@@ -1262,27 +1225,19 @@
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
 ; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm1
-; AVX1-NEXT:    vpackssdw %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX1-NEXT:    # kill: def $al killed $al killed $eax
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmovmskpd %ymm0, %eax
+; AVX1-NEXT:    testb %al, %al
+; AVX1-NEXT:    setne %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: bool_reduction_v4i64:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT:    vpackssdw %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpor %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    # kill: def $al killed $al killed $eax
+; AVX2-NEXT:    vmovmskpd %ymm0, %eax
+; AVX2-NEXT:    testb %al, %al
+; AVX2-NEXT:    setne %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1321,15 +1276,10 @@
 ; SSE-NEXT:    pminud %xmm0, %xmm2
 ; SSE-NEXT:    pcmpeqd %xmm0, %xmm2
 ; SSE-NEXT:    packssdw %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pextrb $0, %xmm0, %eax
-; SSE-NEXT:    # kill: def $al killed $al killed $eax
+; SSE-NEXT:    packsswb %xmm0, %xmm2
+; SSE-NEXT:    pmovmskb %xmm2, %eax
+; SSE-NEXT:    testb %al, %al
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v8i32:
@@ -1340,15 +1290,10 @@
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm3, %xmm2
 ; AVX1-NEXT:    vpminud %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX1-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX1-NEXT:    # kill: def $al killed $al killed $eax
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmovmskps %ymm0, %eax
+; AVX1-NEXT:    testb %al, %al
+; AVX1-NEXT:    setne %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -1356,16 +1301,9 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpminud %ymm1, %ymm0, %ymm1
 ; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX2-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT:    vpor %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX2-NEXT:    # kill: def $al killed $al killed $eax
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    testb %al, %al
+; AVX2-NEXT:    setne %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1410,8 +1348,8 @@
 ; SSE-NEXT:    pcmpeqw %xmm2, %xmm0
 ; SSE-NEXT:    packsswb %xmm1, %xmm0
 ; SSE-NEXT:    pmovmskb %xmm0, %eax
-; SSE-NEXT:    negl %eax
-; SSE-NEXT:    sbbb %al, %al
+; SSE-NEXT:    testw %ax, %ax
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v16i16:
@@ -1422,8 +1360,8 @@
 ; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
-; AVX1-NEXT:    negl %eax
-; AVX1-NEXT:    sbbb %al, %al
+; AVX1-NEXT:    testw %ax, %ax
+; AVX1-NEXT:    setne %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -1433,8 +1371,8 @@
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpmovmskb %xmm0, %eax
-; AVX2-NEXT:    negl %eax
-; AVX2-NEXT:    sbbb %al, %al
+; AVX2-NEXT:    testw %ax, %ax
+; AVX2-NEXT:    setne %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1469,12 +1407,12 @@
 define i1 @bool_reduction_v32i8(<32 x i8> %x, <32 x i8> %y) {
 ; SSE-LABEL: bool_reduction_v32i8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqb %xmm2, %xmm0
 ; SSE-NEXT:    pcmpeqb %xmm3, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pmovmskb %xmm1, %eax
-; SSE-NEXT:    negl %eax
-; SSE-NEXT:    sbbb %al, %al
+; SSE-NEXT:    pcmpeqb %xmm2, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    testw %ax, %ax
+; SSE-NEXT:    setne %al
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: bool_reduction_v32i8:
@@ -1485,8 +1423,8 @@
 ; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpor %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vpmovmskb %xmm0, %eax
-; AVX1-NEXT:    negl %eax
-; AVX1-NEXT:    sbbb %al, %al
+; AVX1-NEXT:    testw %ax, %ax
+; AVX1-NEXT:    setne %al
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
@@ -1494,8 +1432,8 @@
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpmovmskb %ymm0, %eax
-; AVX2-NEXT:    negl %eax
-; AVX2-NEXT:    sbbb %al, %al
+; AVX2-NEXT:    testl %eax, %eax
+; AVX2-NEXT:    setne %al
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;