[X86][SSE] Add support for X86ISD::PACKSS to ComputeNumSignBitsForTargetNode

[RKSimon]

Summary

Helps improve combineLogicBlendIntoPBLENDV support by allowing us to peek into through PACKSS truncations of vector comparison results

[delena]

I want to suggest a test case without intrinsic:
define <16 x i8> @vselect_packss_opt(<16 x i16> %a0, <16 x i16> %a1, <16 x i8> %a2, <16 x i8> %a3) {

%1 = icmp eq <16 x i16> %a0, %a1
%x2 = sext <16 x i1> %1 to <16 x i8>
%x6 = and <16 x i8> %x2, %a2
%x7 = xor <16 x i8> %x2, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
%x8 = and <16 x i8> %x7, %a3
%x9 = or <16 x i8> %x6, %x8
ret <16 x i8> %x9

}

[RKSimon]

Added general test cases (not directly calling PACKSS intrinsic).

I added v16i32/v16i64 cases although these need some further work to be handled in combineLogicBlendIntoPBLENDV

[RKSimon]

In D37680#866273, @delena wrote:

I want to suggest a test case without intrinsic:

Done - I added vselect_packss_v16i32 / vselect_packss_v16i64 cases as well. Interestingly avx512vl gets caught in an infinite loop on vselect_packss_v16i32, with/without this patch - I'll raise a bug.

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

unsigned X86TargetLowering::ComputeNumSignBitsForTargetNode(
 
   case X86ISD::VSEXT: {
     SDValue Src = Op.getOperand(0);
     unsigned Tmp = DAG.ComputeNumSignBits(Src, Depth + 1);
     Tmp += VTBits - Src.getScalarValueSizeInBits();
     return Tmp;
   }
 
+  case X86ISD::PACKSS: {
+    // PACKSS is just a truncation if the sign bits extend to the packed size.
+    // TODO: Add DemandedElts support.
+    unsigned SrcBits = Op.getOperand(0).getScalarValueSizeInBits();
+    unsigned Tmp0 = DAG.ComputeNumSignBits(Op.getOperand(0), Depth + 1);
+    unsigned Tmp1 = DAG.ComputeNumSignBits(Op.getOperand(1), Depth + 1);
+    unsigned Tmp = std::min(Tmp0, Tmp1);
+    if (Tmp > (SrcBits - VTBits))
+      return Tmp - (SrcBits - VTBits);
+    return 1;
+  }
+
   case X86ISD::VSHLI: {
     SDValue Src = Op.getOperand(0);
     unsigned Tmp = DAG.ComputeNumSignBits(Src, Depth + 1);
     APInt ShiftVal = cast<ConstantSDNode>(Op.getOperand(1))->getAPIntValue();
     if (ShiftVal.uge(VTBits))
       return VTBits; // Shifted all bits out --> zero.
     if (ShiftVal.uge(Tmp))
       return 1; // Shifted all sign bits out --> unknown.

llvm/trunk/test/CodeGen/X86/bitcast-and-setcc-512.ll

-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX12,AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX12,AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512BW
-
-define i8 @v8i64(<8 x i64> %a, <8 x i64> %b, <8 x i64> %c, <8 x i64> %d) {
-; SSE-LABEL: v8i64:
-; SSE:       # BB#0:
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pcmpgtq %xmm7, %xmm3
-; SSE-NEXT:    pcmpgtq %xmm6, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
-; SSE-NEXT:    pslld $31, %xmm2
-; SSE-NEXT:    psrad $31, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; SSE-NEXT:    pshufb %xmm3, %xmm2
-; SSE-NEXT:    pcmpgtq %xmm5, %xmm1
-; SSE-NEXT:    pcmpgtq %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    pslld $31, %xmm0
-; SSE-NEXT:    psrad $31, %xmm0
-; SSE-NEXT:    pshufb %xmm3, %xmm0
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE-NEXT:    psllw $15, %xmm0
-; SSE-NEXT:    psraw $15, %xmm0
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm11[0,2]
-; SSE-NEXT:    pslld $31, %xmm9
-; SSE-NEXT:    psrad $31, %xmm9
-; SSE-NEXT:    pshufb %xmm3, %xmm9
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm10[0,2]
-; SSE-NEXT:    pslld $31, %xmm8
-; SSE-NEXT:    psrad $31, %xmm8
-; SSE-NEXT:    pshufb %xmm3, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
-; SSE-NEXT:    psllw $15, %xmm8
-; SSE-NEXT:    psraw $15, %xmm8
-; SSE-NEXT:    pand %xmm0, %xmm8
-; SSE-NEXT:    pshufb {{.*#+}} xmm8 = xmm8[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
-; SSE-NEXT:    pmovmskb %xmm8, %eax
-; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: v8i64:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
-; AVX1-NEXT:    vpcmpgtq %xmm8, %xmm9, %xmm8
-; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpacksswb %xmm8, %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm8 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm9
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb %xmm8, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm9[0]
-; AVX1-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm2
-; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpcmpgtq %xmm7, %xmm5, %xmm2
-; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
-; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtq %xmm6, %xmm4, %xmm3
-; AVX1-NEXT:    vpacksswb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpshufb %xmm8, %xmm2, %xmm2
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
-; AVX1-NEXT:    vpsllw $15, %xmm1, %xmm1
-; AVX1-NEXT:    vpsraw $15, %xmm1, %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: v8i64:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpcmpgtq %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX2-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX2-NEXT:    vpsraw $15, %xmm0, %xmm0
-; AVX2-NEXT:    vpcmpgtq %ymm7, %ymm5, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtq %ymm6, %ymm4, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm4
-; AVX2-NEXT:    vpacksswb %xmm4, %xmm2, %xmm2
-; AVX2-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
-; AVX2-NEXT:    vpsllw $15, %xmm1, %xmm1
-; AVX2-NEXT:    vpsraw $15, %xmm1, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; 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
-;
-; AVX512F-LABEL: v8i64:
-; AVX512F:       # BB#0:
-; AVX512F-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1
-; AVX512F-NEXT:    vpcmpgtq %zmm3, %zmm2, %k0 {%k1}
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: v8i64:
-; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpcmpgtq %zmm3, %zmm2, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-  %x0 = icmp sgt <8 x i64> %a, %b
-  %x1 = icmp sgt <8 x i64> %c, %d
-  %y = and <8 x i1> %x0, %x1
-  %res = bitcast <8 x i1> %y to i8
-  ret i8 %res
-}
-
-define i8 @v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x double> %d) {
-; SSE-LABEL: v8f64:
-; SSE:       # BB#0:
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    cmpltpd %xmm3, %xmm7
-; SSE-NEXT:    cmpltpd %xmm2, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[0,2]
-; SSE-NEXT:    pslld $31, %xmm6
-; SSE-NEXT:    psrad $31, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; SSE-NEXT:    pshufb %xmm2, %xmm6
-; SSE-NEXT:    cmpltpd %xmm1, %xmm5
-; SSE-NEXT:    cmpltpd %xmm0, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm5[0,2]
-; SSE-NEXT:    pslld $31, %xmm4
-; SSE-NEXT:    psrad $31, %xmm4
-; SSE-NEXT:    pshufb %xmm2, %xmm4
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm6[0]
-; SSE-NEXT:    psllw $15, %xmm4
-; SSE-NEXT:    psraw $15, %xmm4
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm11[0,2]
-; SSE-NEXT:    pslld $31, %xmm9
-; SSE-NEXT:    psrad $31, %xmm9
-; SSE-NEXT:    pshufb %xmm2, %xmm9
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm10[0,2]
-; SSE-NEXT:    pslld $31, %xmm8
-; SSE-NEXT:    psrad $31, %xmm8
-; SSE-NEXT:    pshufb %xmm2, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
-; SSE-NEXT:    psllw $15, %xmm8
-; SSE-NEXT:    psraw $15, %xmm8
-; SSE-NEXT:    pand %xmm4, %xmm8
-; SSE-NEXT:    pshufb {{.*#+}} xmm8 = xmm8[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
-; SSE-NEXT:    pmovmskb %xmm8, %eax
-; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; SSE-NEXT:    retq
-;
-; AVX12-LABEL: v8f64:
-; AVX12:       # BB#0:
-; AVX12-NEXT:    vcmpltpd %ymm1, %ymm3, %ymm1
-; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm3
-; AVX12-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
-; AVX12-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX12-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX12-NEXT:    vcmpltpd %ymm0, %ymm2, %ymm0
-; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX12-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX12-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
-; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX12-NEXT:    vpsllw $15, %xmm0, %xmm0
-; AVX12-NEXT:    vpsraw $15, %xmm0, %xmm0
-; AVX12-NEXT:    vcmpltpd %ymm5, %ymm7, %ymm1
-; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX12-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
-; AVX12-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX12-NEXT:    vcmpltpd %ymm4, %ymm6, %ymm2
-; AVX12-NEXT:    vextractf128 $1, %ymm2, %xmm4
-; AVX12-NEXT:    vpacksswb %xmm4, %xmm2, %xmm2
-; AVX12-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
-; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
-; AVX12-NEXT:    vpsllw $15, %xmm1, %xmm1
-; AVX12-NEXT:    vpsraw $15, %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
-;
-; AVX512F-LABEL: v8f64:
-; AVX512F:       # BB#0:
-; AVX512F-NEXT:    vcmpltpd %zmm0, %zmm1, %k1
-; AVX512F-NEXT:    vcmpltpd %zmm2, %zmm3, %k0 {%k1}
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: v8f64:
-; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vcmpltpd %zmm0, %zmm1, %k1
-; AVX512BW-NEXT:    vcmpltpd %zmm2, %zmm3, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-  %x0 = fcmp ogt <8 x double> %a, %b
-  %x1 = fcmp ogt <8 x double> %c, %d
-  %y = and <8 x i1> %x0, %x1
-  %res = bitcast <8 x i1> %y to i8
-  ret i8 %res
-}
-
-define i32 @v32i16(<32 x i16> %a, <32 x i16> %b, <32 x i16> %c, <32 x i16> %d) {
-; SSE-LABEL: v32i16:
-; SSE:       # BB#0:
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pcmpgtw %xmm5, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm5, %xmm1
-; SSE-NEXT:    pcmpgtw %xmm4, %xmm0
-; SSE-NEXT:    pshufb %xmm5, %xmm0
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE-NEXT:    psllw $7, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    pcmpgtb %xmm0, %xmm4
-; SSE-NEXT:    pcmpgtw %xmm7, %xmm3
-; SSE-NEXT:    pshufb %xmm5, %xmm3
-; SSE-NEXT:    pcmpgtw %xmm6, %xmm2
-; SSE-NEXT:    pshufb %xmm5, %xmm2
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
-; SSE-NEXT:    psllw $7, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    pcmpgtb %xmm2, %xmm0
-; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pshufb %xmm5, %xmm11
-; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pshufb %xmm5, %xmm10
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm10 = xmm10[0],xmm11[0]
-; SSE-NEXT:    psllw $7, %xmm10
-; SSE-NEXT:    pand %xmm12, %xmm10
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    pcmpgtb %xmm10, %xmm2
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pshufb %xmm5, %xmm9
-; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pshufb %xmm5, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
-; SSE-NEXT:    psllw $7, %xmm8
-; SSE-NEXT:    pand %xmm12, %xmm8
-; SSE-NEXT:    pcmpgtb %xmm8, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pmovmskb %xmm2, %ecx
-; SSE-NEXT:    pmovmskb %xmm1, %eax
-; SSE-NEXT:    shll $16, %eax
-; SSE-NEXT:    orl %ecx, %eax
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: v32i16:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
-; AVX1-NEXT:    vpcmpgtw %xmm8, %xmm9, %xmm8
-; AVX1-NEXT:    vpcmpgtw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpacksswb %xmm8, %xmm1, %xmm8
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpgtw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpcmpgtw %xmm7, %xmm5, %xmm2
-; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm8, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
-; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtw %xmm6, %xmm4, %xmm3
-; AVX1-NEXT:    vpacksswb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpsllw $7, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
-; AVX1-NEXT:    vpsllw $7, %xmm1, %xmm0
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm3, %xmm0
-; AVX1-NEXT:    vpmovmskb %xmm0, %eax
-; AVX1-NEXT:    shll $16, %eax
-; AVX1-NEXT:    orl %ecx, %eax
-; AVX1-NEXT:    vzeroupper
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: v32i16:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpcmpgtw %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpcmpgtw %ymm7, %ymm5, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtw %ymm6, %ymm4, %ymm2
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm2, %xmm2
-; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpsllw $7, %ymm0, %ymm0
-; AVX2-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpmovmskb %ymm0, %eax
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512F-LABEL: v32i16:
-; AVX512F:       # BB#0:
-; AVX512F-NEXT:    pushq %rbp
-; AVX512F-NEXT:  .Lcfi0:
-; AVX512F-NEXT:    .cfi_def_cfa_offset 16
-; AVX512F-NEXT:  .Lcfi1:
-; AVX512F-NEXT:    .cfi_offset %rbp, -16
-; AVX512F-NEXT:    movq %rsp, %rbp
-; AVX512F-NEXT:  .Lcfi2:
-; AVX512F-NEXT:    .cfi_def_cfa_register %rbp
-; AVX512F-NEXT:    andq $-32, %rsp
-; AVX512F-NEXT:    subq $32, %rsp
-; AVX512F-NEXT:    vpcmpgtw %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpmovsxwd %ymm1, %zmm1
-; AVX512F-NEXT:    vpslld $31, %zmm1, %zmm1
-; AVX512F-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %ecx
-; AVX512F-NEXT:    vmovd %ecx, %xmm1
-; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm0
-; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
-; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %ecx
-; AVX512F-NEXT:    vmovd %ecx, %xmm0
-; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
-; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512F-NEXT:    vpcmpgtw %ymm7, %ymm5, %ymm1
-; AVX512F-NEXT:    vpmovsxwd %ymm1, %zmm1
-; AVX512F-NEXT:    vpslld $31, %zmm1, %zmm1
-; AVX512F-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %ecx
-; AVX512F-NEXT:    vmovd %ecx, %xmm1
-; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm1
-; AVX512F-NEXT:    vpcmpgtw %ymm6, %ymm4, %ymm2
-; AVX512F-NEXT:    vpmovsxwd %ymm2, %zmm2
-; AVX512F-NEXT:    vpslld $31, %zmm2, %zmm2
-; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %ecx
-; AVX512F-NEXT:    vmovd %ecx, %xmm2
-; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
-; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
-; AVX512F-NEXT:    kmovw %k1, %eax
-; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm2, %xmm2
-; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
-; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX512F-NEXT:    vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT:    vpslld $31, %zmm1, %zmm1
-; AVX512F-NEXT:    vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
-; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
-; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
-; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; AVX512F-NEXT:    kmovw %k0, (%rsp)
-; AVX512F-NEXT:    movl (%rsp), %eax
-; AVX512F-NEXT:    movq %rbp, %rsp
-; AVX512F-NEXT:    popq %rbp
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: v32i16:
-; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpcmpgtw %zmm3, %zmm2, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-  %x0 = icmp sgt <32 x i16> %a, %b
-  %x1 = icmp sgt <32 x i16> %c, %d
-  %y = and <32 x i1> %x0, %x1
-  %res = bitcast <32 x i1> %y to i32
-  ret i32 %res
-}
-
-define i16 @v16i32(<16 x i32> %a, <16 x i32> %b, <16 x i32> %c, <16 x i32> %d) {
-; SSE-LABEL: v16i32:
-; SSE:       # BB#0:
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pcmpgtd %xmm7, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; SSE-NEXT:    pshufb %xmm7, %xmm3
-; SSE-NEXT:    pcmpgtd %xmm6, %xmm2
-; SSE-NEXT:    pshufb %xmm7, %xmm2
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
-; SSE-NEXT:    psllw $15, %xmm2
-; SSE-NEXT:    psraw $15, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm3, %xmm2
-; SSE-NEXT:    pcmpgtd %xmm5, %xmm1
-; SSE-NEXT:    pshufb %xmm7, %xmm1
-; SSE-NEXT:    pcmpgtd %xmm4, %xmm0
-; SSE-NEXT:    pshufb %xmm7, %xmm0
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE-NEXT:    psllw $15, %xmm0
-; SSE-NEXT:    psraw $15, %xmm0
-; SSE-NEXT:    pshufb %xmm3, %xmm0
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE-NEXT:    psllw $7, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    pcmpgtb %xmm0, %xmm4
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pshufb %xmm7, %xmm11
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pshufb %xmm7, %xmm9
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm11[0]
-; SSE-NEXT:    psllw $15, %xmm9
-; SSE-NEXT:    psraw $15, %xmm9
-; SSE-NEXT:    pshufb %xmm3, %xmm9
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pshufb %xmm7, %xmm10
-; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pshufb %xmm7, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm10[0]
-; SSE-NEXT:    psllw $15, %xmm8
-; SSE-NEXT:    psraw $15, %xmm8
-; SSE-NEXT:    pshufb %xmm3, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
-; SSE-NEXT:    psllw $7, %xmm8
-; SSE-NEXT:    pand %xmm2, %xmm8
-; SSE-NEXT:    pcmpgtb %xmm8, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    pmovmskb %xmm1, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: v16i32:
-; AVX1:       # BB#0:
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
-; AVX1-NEXT:    vpcmpgtd %xmm8, %xmm9, %xmm8
-; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpacksswb %xmm8, %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm8 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm9
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpshufb %xmm8, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm9[0]
-; AVX1-NEXT:    vpsllw $7, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm9 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX1-NEXT:    vpand %xmm9, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm1
-; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtd %xmm7, %xmm5, %xmm3
-; AVX1-NEXT:    vpacksswb %xmm1, %xmm3, %xmm1
-; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm1
-; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm5
-; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm5, %xmm3
-; AVX1-NEXT:    vpcmpgtd %xmm6, %xmm4, %xmm4
-; AVX1-NEXT:    vpacksswb %xmm3, %xmm4, %xmm3
-; AVX1-NEXT:    vpshufb %xmm8, %xmm3, %xmm3
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm3[0],xmm1[0]
-; AVX1-NEXT:    vpsllw $7, %xmm1, %xmm1
-; AVX1-NEXT:    vpand %xmm9, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtb %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: v16i32:
-; AVX2:       # BB#0:
-; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
-; AVX2-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpgtd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX2-NEXT:    vpsllw $7, %xmm0, %xmm0
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
-; AVX2-NEXT:    vpcmpgtd %ymm7, %ymm5, %ymm5
-; AVX2-NEXT:    vextracti128 $1, %ymm5, %xmm7
-; AVX2-NEXT:    vpacksswb %xmm7, %xmm5, %xmm5
-; AVX2-NEXT:    vpshufb %xmm3, %xmm5, %xmm5
-; AVX2-NEXT:    vpcmpgtd %ymm6, %ymm4, %ymm4
-; AVX2-NEXT:    vextracti128 $1, %ymm4, %xmm6
-; AVX2-NEXT:    vpacksswb %xmm6, %xmm4, %xmm4
-; AVX2-NEXT:    vpshufb %xmm3, %xmm4, %xmm3
-; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
-; AVX2-NEXT:    vpsllw $7, %xmm3, %xmm3
-; AVX2-NEXT:    vpand %xmm1, %xmm3, %xmm1
-; AVX2-NEXT:    vpcmpgtb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
-; AVX2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512F-LABEL: v16i32:
-; AVX512F:       # BB#0:
-; AVX512F-NEXT:    vpcmpgtd %zmm1, %zmm0, %k1
-; AVX512F-NEXT:    vpcmpgtd %zmm3, %zmm2, %k0 {%k1}
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: v16i32:
-; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpcmpgtd %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpcmpgtd %zmm3, %zmm2, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-  %x0 = icmp sgt <16 x i32> %a, %b
-  %x1 = icmp sgt <16 x i32> %c, %d
-  %y = and <16 x i1> %x0, %x1
-  %res = bitcast <16 x i1> %y to i16
-  ret i16 %res
-}
-
-define i16 @v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x float> %d) {
-; SSE-LABEL: v16f32:
-; SSE:       # BB#0:
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    cmpltps %xmm3, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; SSE-NEXT:    pshufb %xmm3, %xmm7
-; SSE-NEXT:    cmpltps %xmm2, %xmm6
-; SSE-NEXT:    pshufb %xmm3, %xmm6
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm7[0]
-; SSE-NEXT:    psllw $15, %xmm6
-; SSE-NEXT:    psraw $15, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; SSE-NEXT:    pshufb %xmm2, %xmm6
-; SSE-NEXT:    cmpltps %xmm1, %xmm5
-; SSE-NEXT:    pshufb %xmm3, %xmm5
-; SSE-NEXT:    cmpltps %xmm0, %xmm4
-; SSE-NEXT:    pshufb %xmm3, %xmm4
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm5[0]
-; SSE-NEXT:    psllw $15, %xmm4
-; SSE-NEXT:    psraw $15, %xmm4
-; SSE-NEXT:    pshufb %xmm2, %xmm4
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm6[0]
-; SSE-NEXT:    psllw $7, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; SSE-NEXT:    pand %xmm1, %xmm4
-; SSE-NEXT:    xorps %xmm0, %xmm0
-; SSE-NEXT:    pxor %xmm5, %xmm5
-; SSE-NEXT:    pcmpgtb %xmm4, %xmm5
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pshufb %xmm3, %xmm11
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pshufb %xmm3, %xmm9
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm11[0]
-; SSE-NEXT:    psllw $15, %xmm9
-; SSE-NEXT:    psraw $15, %xmm9
-; SSE-NEXT:    pshufb %xmm2, %xmm9
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pshufb %xmm3, %xmm10
-; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pshufb %xmm3, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm10[0]
-; SSE-NEXT:    psllw $15, %xmm8
-; SSE-NEXT:    psraw $15, %xmm8
-; SSE-NEXT:    pshufb %xmm2, %xmm8
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
-; SSE-NEXT:    psllw $7, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm8
-; SSE-NEXT:    pcmpgtb %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    pmovmskb %xmm0, %eax
-; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; SSE-NEXT:    retq
-;
-; AVX12-LABEL: v16f32:
-; AVX12:       # BB#0:
-; AVX12-NEXT:    vcmpltps %ymm1, %ymm3, %ymm1
-; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm3
-; AVX12-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
-; AVX12-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
-; AVX12-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
-; AVX12-NEXT:    vcmpltps %ymm0, %ymm2, %ymm0
-; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm2
-; AVX12-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
-; AVX12-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
-; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX12-NEXT:    vpsllw $7, %xmm0, %xmm0
-; AVX12-NEXT:    vmovdqa {{.*#+}} xmm1 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX12-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
-; AVX12-NEXT:    vcmpltps %ymm5, %ymm7, %ymm5
-; AVX12-NEXT:    vextractf128 $1, %ymm5, %xmm7
-; AVX12-NEXT:    vpacksswb %xmm7, %xmm5, %xmm5
-; AVX12-NEXT:    vpshufb %xmm3, %xmm5, %xmm5
-; AVX12-NEXT:    vcmpltps %ymm4, %ymm6, %ymm4
-; AVX12-NEXT:    vextractf128 $1, %ymm4, %xmm6
-; AVX12-NEXT:    vpacksswb %xmm6, %xmm4, %xmm4
-; AVX12-NEXT:    vpshufb %xmm3, %xmm4, %xmm3
-; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
-; AVX12-NEXT:    vpsllw $7, %xmm3, %xmm3
-; AVX12-NEXT:    vpand %xmm1, %xmm3, %xmm1
-; AVX12-NEXT:    vpcmpgtb %xmm1, %xmm2, %xmm1
-; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX12-NEXT:    vpmovmskb %xmm0, %eax
-; AVX12-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; AVX12-NEXT:    vzeroupper
-; AVX12-NEXT:    retq
-;
-; AVX512F-LABEL: v16f32:
-; AVX512F:       # BB#0:
-; AVX512F-NEXT:    vcmpltps %zmm0, %zmm1, %k1
-; AVX512F-NEXT:    vcmpltps %zmm2, %zmm3, %k0 {%k1}
-; AVX512F-NEXT:    kmovw %k0, %eax
-; AVX512F-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: v16f32:
-; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vcmpltps %zmm0, %zmm1, %k1
-; AVX512BW-NEXT:    vcmpltps %zmm2, %zmm3, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-  %x0 = fcmp ogt <16 x float> %a, %b
-  %x1 = fcmp ogt <16 x float> %c, %d
-  %y = and <16 x i1> %x0, %x1
-  %res = bitcast <16 x i1> %y to i16
-  ret i16 %res
-}
-
-define i64 @v64i8(<64 x i8> %a, <64 x i8> %b, <64 x i8> %c, <64 x i8> %d) {
-; SSE-LABEL: v64i8:
-; SSE:       # BB#0:
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pcmpgtb %xmm6, %xmm2
-; SSE-NEXT:    pcmpgtb %xmm7, %xmm3
-; SSE-NEXT:    pcmpgtb %xmm4, %xmm0
-; SSE-NEXT:    pcmpgtb %xmm5, %xmm1
-; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm8
-; SSE-NEXT:    pand %xmm2, %xmm8
-; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm9
-; SSE-NEXT:    pand %xmm3, %xmm9
-; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm10
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm11
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    pextrb $15, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $14, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $13, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $12, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $11, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $10, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $9, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $8, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $7, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $6, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $5, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $4, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $3, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $2, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $1, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $0, %xmm11, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $15, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $14, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $13, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $12, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $11, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $10, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $9, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $8, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $7, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $6, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $5, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $4, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $3, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $2, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $1, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $0, %xmm10, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $15, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $14, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $13, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $12, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $11, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $10, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $9, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $8, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $7, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $6, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $5, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $4, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $3, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $2, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $1, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $0, %xmm9, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $15, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $14, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $13, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $12, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $11, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $10, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $9, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $8, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $7, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $6, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $5, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $4, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $3, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $2, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $1, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    pextrb $0, %xmm8, %eax
-; SSE-NEXT:    andb $1, %al
-; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
-; SSE-NEXT:    movl -{{[0-9]+}}(%rsp), %eax
-; SSE-NEXT:    shll $16, %eax
-; SSE-NEXT:    movzwl -{{[0-9]+}}(%rsp), %ecx
-; SSE-NEXT:    orl %eax, %ecx
-; SSE-NEXT:    movl -{{[0-9]+}}(%rsp), %edx
-; SSE-NEXT:    shll $16, %edx
-; SSE-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
-; SSE-NEXT:    orl %edx, %eax
-; SSE-NEXT:    shlq $32, %rax
-; SSE-NEXT:    orq %rcx, %rax
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: v64i8:
-; 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 $64, %rsp
-; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
-; AVX1-NEXT:    vpcmpgtb %xmm8, %xmm9, %xmm8
-; AVX1-NEXT:    vpcmpgtb %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vinsertf128 $1, %xmm8, %ymm1, %ymm8
-; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpgtb %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpgtb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm0
-; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm2
-; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
-; AVX1-NEXT:    vpcmpgtb %xmm7, %xmm5, %xmm2
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-NEXT:    vandps %ymm0, %ymm8, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
-; AVX1-NEXT:    vpcmpgtb %xmm2, %xmm3, %xmm2
-; AVX1-NEXT:    vpcmpgtb %xmm6, %xmm4, %xmm3
-; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
-; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm1
-; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
-; AVX1-NEXT:    vpextrb $15, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $14, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $13, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $12, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $11, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $10, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $9, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $8, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $7, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $6, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $5, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $4, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $3, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $2, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $1, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; AVX1-NEXT:    vpextrb $0, %xmm2, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, (%rsp)
-; 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:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpextrb $15, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $14, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $13, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $12, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $11, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $10, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $9, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $8, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $7, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $6, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $5, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $4, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $3, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $2, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $1, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX1-NEXT:    vpextrb $0, %xmm1, %eax
-; AVX1-NEXT:    andb $1, %al
-; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; 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:    movl (%rsp), %ecx
-; AVX1-NEXT:    movl {{[0-9]+}}(%rsp), %eax
-; AVX1-NEXT:    shlq $32, %rax
-; AVX1-NEXT:    orq %rcx, %rax
-; AVX1-NEXT:    movq %rbp, %rsp
-; AVX1-NEXT:    popq %rbp
-; AVX1-NEXT:    vzeroupper
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: v64i8:
-; 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 $64, %rsp
-; AVX2-NEXT:    vpcmpgtb %ymm3, %ymm1, %ymm1
-; AVX2-NEXT:    vpcmpgtb %ymm2, %ymm0, %ymm2
-; AVX2-NEXT:    vpcmpgtb %ymm7, %ymm5, %ymm0
-; AVX2-NEXT:    vpand %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vpcmpgtb %ymm6, %ymm4, %ymm1
-; AVX2-NEXT:    vpand %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
-; AVX2-NEXT:    vpextrb $15, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm2, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, (%rsp)
-; 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:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpextrb $15, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $14, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $13, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $12, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $11, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $10, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $9, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $8, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $7, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $6, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $5, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $4, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $3, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $2, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $1, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; AVX2-NEXT:    vpextrb $0, %xmm1, %eax
-; AVX2-NEXT:    andb $1, %al
-; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
-; 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:    movl (%rsp), %ecx
-; AVX2-NEXT:    movl {{[0-9]+}}(%rsp), %eax
-; AVX2-NEXT:    shlq $32, %rax
-; AVX2-NEXT:    orq %rcx, %rax
-; AVX2-NEXT:    movq %rbp, %rsp
-; AVX2-NEXT:    popq %rbp
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512F-LABEL: v64i8:
-; AVX512F:       # BB#0:
-; AVX512F-NEXT:    pushq %rbp
-; AVX512F-NEXT:  .Lcfi3:
-; AVX512F-NEXT:    .cfi_def_cfa_offset 16
-; AVX512F-NEXT:  .Lcfi4:
-; AVX512F-NEXT:    .cfi_offset %rbp, -16
-; AVX512F-NEXT:    movq %rsp, %rbp
-; AVX512F-NEXT:  .Lcfi5:
-; AVX512F-NEXT:    .cfi_def_cfa_register %rbp
-; AVX512F-NEXT:    andq $-32, %rsp
-; AVX512F-NEXT:    subq $64, %rsp
-; AVX512F-NEXT:    vpcmpgtb %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpcmpgtb %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpcmpgtb %ymm7, %ymm5, %ymm2
-; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpcmpgtb %ymm6, %ymm4, %ymm2
-; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm2
-; AVX512F-NEXT:    vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT:    vpslld $31, %zmm2, %zmm2
-; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
-; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
-; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
-; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; AVX512F-NEXT:    kmovw %k0, (%rsp)
-; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm0
-; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
-; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
-; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
-; AVX512F-NEXT:    vpmovsxbd %xmm1, %zmm0
-; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
-; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
-; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
-; AVX512F-NEXT:    movl (%rsp), %ecx
-; AVX512F-NEXT:    movl {{[0-9]+}}(%rsp), %eax
-; AVX512F-NEXT:    shlq $32, %rax
-; AVX512F-NEXT:    orq %rcx, %rax
-; AVX512F-NEXT:    movq %rbp, %rsp
-; AVX512F-NEXT:    popq %rbp
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: v64i8:
-; AVX512BW:       # BB#0:
-; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpcmpgtb %zmm3, %zmm2, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
-  %x0 = icmp sgt <64 x i8> %a, %b
-  %x1 = icmp sgt <64 x i8> %c, %d
-  %y = and <64 x i1> %x0, %x1
-  %res = bitcast <64 x i1> %y to i64
-  ret i64 %res
-}
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX12,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX12,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512BW
+
+define i8 @v8i64(<8 x i64> %a, <8 x i64> %b, <8 x i64> %c, <8 x i64> %d) {
+; SSE-LABEL: v8i64:
+; SSE:       # BB#0:
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pcmpgtq %xmm7, %xmm3
+; SSE-NEXT:    pcmpgtq %xmm6, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
+; SSE-NEXT:    pslld $31, %xmm2
+; SSE-NEXT:    psrad $31, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE-NEXT:    pshufb %xmm3, %xmm2
+; SSE-NEXT:    pcmpgtq %xmm5, %xmm1
+; SSE-NEXT:    pcmpgtq %xmm4, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; SSE-NEXT:    pslld $31, %xmm0
+; SSE-NEXT:    psrad $31, %xmm0
+; SSE-NEXT:    pshufb %xmm3, %xmm0
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT:    psllw $15, %xmm0
+; SSE-NEXT:    psraw $15, %xmm0
+; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm11[0,2]
+; SSE-NEXT:    pslld $31, %xmm9
+; SSE-NEXT:    psrad $31, %xmm9
+; SSE-NEXT:    pshufb %xmm3, %xmm9
+; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    pcmpgtq {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm10[0,2]
+; SSE-NEXT:    pslld $31, %xmm8
+; SSE-NEXT:    psrad $31, %xmm8
+; SSE-NEXT:    pshufb %xmm3, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
+; SSE-NEXT:    psllw $15, %xmm8
+; SSE-NEXT:    psraw $15, %xmm8
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    pshufb {{.*#+}} xmm8 = xmm8[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSE-NEXT:    pmovmskb %xmm8, %eax
+; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: v8i64:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
+; AVX1-NEXT:    vpcmpgtq %xmm8, %xmm9, %xmm8
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpacksswb %xmm8, %xmm1, %xmm1
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm8 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm9
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpgtq %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm8, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm9[0]
+; AVX1-NEXT:    vpsllw $15, %xmm0, %xmm0
+; AVX1-NEXT:    vpsraw $15, %xmm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm2
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpgtq %xmm7, %xmm5, %xmm2
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpcmpgtq %xmm6, %xmm4, %xmm3
+; AVX1-NEXT:    vpacksswb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpshufb %xmm8, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-NEXT:    vpsllw $15, %xmm1, %xmm1
+; AVX1-NEXT:    vpsraw $15, %xmm1, %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: v8i64:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vpcmpgtq %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; AVX2-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtq %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-NEXT:    vpsllw $15, %xmm0, %xmm0
+; AVX2-NEXT:    vpsraw $15, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq %ymm7, %ymm5, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
+; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtq %ymm6, %ymm4, %ymm2
+; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm4
+; AVX2-NEXT:    vpacksswb %xmm4, %xmm2, %xmm2
+; AVX2-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX2-NEXT:    vpsllw $15, %xmm1, %xmm1
+; AVX2-NEXT:    vpsraw $15, %xmm1, %xmm1
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; 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
+;
+; AVX512F-LABEL: v8i64:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1
+; AVX512F-NEXT:    vpcmpgtq %zmm3, %zmm2, %k0 {%k1}
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: v8i64:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpcmpgtq %zmm3, %zmm2, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %x0 = icmp sgt <8 x i64> %a, %b
+  %x1 = icmp sgt <8 x i64> %c, %d
+  %y = and <8 x i1> %x0, %x1
+  %res = bitcast <8 x i1> %y to i8
+  ret i8 %res
+}
+
+define i8 @v8f64(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x double> %d) {
+; SSE-LABEL: v8f64:
+; SSE:       # BB#0:
+; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    movapd {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    cmpltpd %xmm3, %xmm7
+; SSE-NEXT:    cmpltpd %xmm2, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[0,2]
+; SSE-NEXT:    pslld $31, %xmm6
+; SSE-NEXT:    psrad $31, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE-NEXT:    pshufb %xmm2, %xmm6
+; SSE-NEXT:    cmpltpd %xmm1, %xmm5
+; SSE-NEXT:    cmpltpd %xmm0, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm5[0,2]
+; SSE-NEXT:    pslld $31, %xmm4
+; SSE-NEXT:    psrad $31, %xmm4
+; SSE-NEXT:    pshufb %xmm2, %xmm4
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    psllw $15, %xmm4
+; SSE-NEXT:    psraw $15, %xmm4
+; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm11[0,2]
+; SSE-NEXT:    pslld $31, %xmm9
+; SSE-NEXT:    psrad $31, %xmm9
+; SSE-NEXT:    pshufb %xmm2, %xmm9
+; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    cmpltpd {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm10[0,2]
+; SSE-NEXT:    pslld $31, %xmm8
+; SSE-NEXT:    psrad $31, %xmm8
+; SSE-NEXT:    pshufb %xmm2, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
+; SSE-NEXT:    psllw $15, %xmm8
+; SSE-NEXT:    psraw $15, %xmm8
+; SSE-NEXT:    pand %xmm4, %xmm8
+; SSE-NEXT:    pshufb {{.*#+}} xmm8 = xmm8[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]
+; SSE-NEXT:    pmovmskb %xmm8, %eax
+; SSE-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; SSE-NEXT:    retq
+;
+; AVX12-LABEL: v8f64:
+; AVX12:       # BB#0:
+; AVX12-NEXT:    vcmpltpd %ymm1, %ymm3, %ymm1
+; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX12-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
+; AVX12-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; AVX12-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX12-NEXT:    vcmpltpd %ymm0, %ymm2, %ymm0
+; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX12-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
+; AVX12-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX12-NEXT:    vpsllw $15, %xmm0, %xmm0
+; AVX12-NEXT:    vpsraw $15, %xmm0, %xmm0
+; AVX12-NEXT:    vcmpltpd %ymm5, %ymm7, %ymm1
+; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX12-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
+; AVX12-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX12-NEXT:    vcmpltpd %ymm4, %ymm6, %ymm2
+; AVX12-NEXT:    vextractf128 $1, %ymm2, %xmm4
+; AVX12-NEXT:    vpacksswb %xmm4, %xmm2, %xmm2
+; AVX12-NEXT:    vpshufb %xmm3, %xmm2, %xmm2
+; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX12-NEXT:    vpsllw $15, %xmm1, %xmm1
+; AVX12-NEXT:    vpsraw $15, %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
+;
+; AVX512F-LABEL: v8f64:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vcmpltpd %zmm0, %zmm1, %k1
+; AVX512F-NEXT:    vcmpltpd %zmm2, %zmm3, %k0 {%k1}
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: v8f64:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vcmpltpd %zmm0, %zmm1, %k1
+; AVX512BW-NEXT:    vcmpltpd %zmm2, %zmm3, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %x0 = fcmp ogt <8 x double> %a, %b
+  %x1 = fcmp ogt <8 x double> %c, %d
+  %y = and <8 x i1> %x0, %x1
+  %res = bitcast <8 x i1> %y to i8
+  ret i8 %res
+}
+
+define i32 @v32i16(<32 x i16> %a, <32 x i16> %b, <32 x i16> %c, <32 x i16> %d) {
+; SSE-LABEL: v32i16:
+; SSE:       # BB#0:
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pcmpgtw %xmm5, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; SSE-NEXT:    pshufb %xmm5, %xmm1
+; SSE-NEXT:    pcmpgtw %xmm4, %xmm0
+; SSE-NEXT:    pshufb %xmm5, %xmm0
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    psllw $7, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    pcmpgtb %xmm0, %xmm4
+; SSE-NEXT:    pcmpgtw %xmm7, %xmm3
+; SSE-NEXT:    pshufb %xmm5, %xmm3
+; SSE-NEXT:    pcmpgtw %xmm6, %xmm2
+; SSE-NEXT:    pshufb %xmm5, %xmm2
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT:    psllw $7, %xmm2
+; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pcmpgtb %xmm2, %xmm0
+; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pshufb %xmm5, %xmm11
+; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    pshufb %xmm5, %xmm10
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm10 = xmm10[0],xmm11[0]
+; SSE-NEXT:    psllw $7, %xmm10
+; SSE-NEXT:    pand %xmm12, %xmm10
+; SSE-NEXT:    pxor %xmm2, %xmm2
+; SSE-NEXT:    pcmpgtb %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    pshufb %xmm5, %xmm9
+; SSE-NEXT:    pcmpgtw {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    pshufb %xmm5, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
+; SSE-NEXT:    psllw $7, %xmm8
+; SSE-NEXT:    pand %xmm12, %xmm8
+; SSE-NEXT:    pcmpgtb %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pmovmskb %xmm2, %ecx
+; SSE-NEXT:    pmovmskb %xmm1, %eax
+; SSE-NEXT:    shll $16, %eax
+; SSE-NEXT:    orl %ecx, %eax
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: v32i16:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
+; AVX1-NEXT:    vpcmpgtw %xmm8, %xmm9, %xmm8
+; AVX1-NEXT:    vpcmpgtw %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpacksswb %xmm8, %xmm1, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpgtw %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm2
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpgtw %xmm7, %xmm5, %xmm2
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpand %xmm1, %xmm8, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpcmpgtw %xmm6, %xmm4, %xmm3
+; AVX1-NEXT:    vpacksswb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovmskb %xmm0, %ecx
+; AVX1-NEXT:    vpmovmskb %xmm1, %eax
+; AVX1-NEXT:    shll $16, %eax
+; AVX1-NEXT:    orl %ecx, %eax
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: v32i16:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vpcmpgtw %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; AVX2-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw %ymm7, %ymm5, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpacksswb %xmm2, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtw %ymm6, %ymm4, %ymm2
+; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm3
+; AVX2-NEXT:    vpacksswb %xmm3, %xmm2, %xmm2
+; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: v32i16:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    pushq %rbp
+; AVX512F-NEXT:  .Lcfi0:
+; AVX512F-NEXT:    .cfi_def_cfa_offset 16
+; AVX512F-NEXT:  .Lcfi1:
+; AVX512F-NEXT:    .cfi_offset %rbp, -16
+; AVX512F-NEXT:    movq %rsp, %rbp
+; AVX512F-NEXT:  .Lcfi2:
+; AVX512F-NEXT:    .cfi_def_cfa_register %rbp
+; AVX512F-NEXT:    andq $-32, %rsp
+; AVX512F-NEXT:    subq $32, %rsp
+; AVX512F-NEXT:    vpcmpgtw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpmovsxwd %ymm1, %zmm1
+; AVX512F-NEXT:    vpslld $31, %zmm1, %zmm1
+; AVX512F-NEXT:    vptestmd %zmm1, %zmm1, %k0
+; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %ecx
+; AVX512F-NEXT:    vmovd %ecx, %xmm1
+; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmovsxwd %ymm0, %zmm0
+; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %ecx
+; AVX512F-NEXT:    vmovd %ecx, %xmm0
+; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
+; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm7, %ymm5, %ymm1
+; AVX512F-NEXT:    vpmovsxwd %ymm1, %zmm1
+; AVX512F-NEXT:    vpslld $31, %zmm1, %zmm1
+; AVX512F-NEXT:    vptestmd %zmm1, %zmm1, %k0
+; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %ecx
+; AVX512F-NEXT:    vmovd %ecx, %xmm1
+; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm1
+; AVX512F-NEXT:    vpcmpgtw %ymm6, %ymm4, %ymm2
+; AVX512F-NEXT:    vpmovsxwd %ymm2, %zmm2
+; AVX512F-NEXT:    vpslld $31, %zmm2, %zmm2
+; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k0
+; AVX512F-NEXT:    kshiftlw $14, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    kshiftlw $15, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %ecx
+; AVX512F-NEXT:    vmovd %ecx, %xmm2
+; AVX512F-NEXT:    vpinsrb $1, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $13, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $2, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $12, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $3, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $11, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $4, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $10, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $5, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $9, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $6, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $8, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $7, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $7, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $8, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $6, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $9, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $5, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $10, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $4, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $11, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $3, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $12, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $2, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $13, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftlw $1, %k0, %k1
+; AVX512F-NEXT:    kshiftrw $15, %k1, %k1
+; AVX512F-NEXT:    kmovw %k1, %eax
+; AVX512F-NEXT:    vpinsrb $14, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    kshiftrw $15, %k0, %k0
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    vpinsrb $15, %eax, %xmm2, %xmm2
+; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX512F-NEXT:    vpmovsxbd %xmm1, %zmm1
+; AVX512F-NEXT:    vpslld $31, %zmm1, %zmm1
+; AVX512F-NEXT:    vptestmd %zmm1, %zmm1, %k0
+; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
+; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
+; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; AVX512F-NEXT:    kmovw %k0, (%rsp)
+; AVX512F-NEXT:    movl (%rsp), %eax
+; AVX512F-NEXT:    movq %rbp, %rsp
+; AVX512F-NEXT:    popq %rbp
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: v32i16:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpcmpgtw %zmm3, %zmm2, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %x0 = icmp sgt <32 x i16> %a, %b
+  %x1 = icmp sgt <32 x i16> %c, %d
+  %y = and <32 x i1> %x0, %x1
+  %res = bitcast <32 x i1> %y to i32
+  ret i32 %res
+}
+
+define i16 @v16i32(<16 x i32> %a, <16 x i32> %b, <16 x i32> %c, <16 x i32> %d) {
+; SSE-LABEL: v16i32:
+; SSE:       # BB#0:
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pcmpgtd %xmm7, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE-NEXT:    pshufb %xmm7, %xmm3
+; SSE-NEXT:    pcmpgtd %xmm6, %xmm2
+; SSE-NEXT:    pshufb %xmm7, %xmm2
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT:    psllw $15, %xmm2
+; SSE-NEXT:    psraw $15, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; SSE-NEXT:    pshufb %xmm3, %xmm2
+; SSE-NEXT:    pcmpgtd %xmm5, %xmm1
+; SSE-NEXT:    pshufb %xmm7, %xmm1
+; SSE-NEXT:    pcmpgtd %xmm4, %xmm0
+; SSE-NEXT:    pshufb %xmm7, %xmm0
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    psllw $15, %xmm0
+; SSE-NEXT:    psraw $15, %xmm0
+; SSE-NEXT:    pshufb %xmm3, %xmm0
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT:    psllw $7, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    pcmpgtb %xmm0, %xmm4
+; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pshufb %xmm7, %xmm11
+; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    pshufb %xmm7, %xmm9
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm11[0]
+; SSE-NEXT:    psllw $15, %xmm9
+; SSE-NEXT:    psraw $15, %xmm9
+; SSE-NEXT:    pshufb %xmm3, %xmm9
+; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    pshufb %xmm7, %xmm10
+; SSE-NEXT:    pcmpgtd {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    pshufb %xmm7, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm10[0]
+; SSE-NEXT:    psllw $15, %xmm8
+; SSE-NEXT:    psraw $15, %xmm8
+; SSE-NEXT:    pshufb %xmm3, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
+; SSE-NEXT:    psllw $7, %xmm8
+; SSE-NEXT:    pand %xmm2, %xmm8
+; SSE-NEXT:    pcmpgtb %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm4, %xmm1
+; SSE-NEXT:    pmovmskb %xmm1, %eax
+; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: v16i32:
+; AVX1:       # BB#0:
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
+; AVX1-NEXT:    vpcmpgtd %xmm8, %xmm9, %xmm8
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpacksswb %xmm8, %xmm1, %xmm1
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm8 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm9
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm8, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm9[0]
+; AVX1-NEXT:    vpsllw $7, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm9 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX1-NEXT:    vpand %xmm9, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm1
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpgtd %xmm7, %xmm5, %xmm3
+; AVX1-NEXT:    vpacksswb %xmm1, %xmm3, %xmm1
+; AVX1-NEXT:    vpshufb %xmm8, %xmm1, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm5
+; AVX1-NEXT:    vpcmpgtd %xmm3, %xmm5, %xmm3
+; AVX1-NEXT:    vpcmpgtd %xmm6, %xmm4, %xmm4
+; AVX1-NEXT:    vpacksswb %xmm3, %xmm4, %xmm3
+; AVX1-NEXT:    vpshufb %xmm8, %xmm3, %xmm3
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm3[0],xmm1[0]
+; AVX1-NEXT:    vpsllw $7, %xmm1, %xmm1
+; AVX1-NEXT:    vpand %xmm9, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpgtb %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: v16i32:
+; AVX2:       # BB#0:
+; AVX2-NEXT:    vpcmpgtd %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; AVX2-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; AVX2-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpgtd %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-NEXT:    vpsllw $7, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
+; AVX2-NEXT:    vpcmpgtd %ymm7, %ymm5, %ymm5
+; AVX2-NEXT:    vextracti128 $1, %ymm5, %xmm7
+; AVX2-NEXT:    vpacksswb %xmm7, %xmm5, %xmm5
+; AVX2-NEXT:    vpshufb %xmm3, %xmm5, %xmm5
+; AVX2-NEXT:    vpcmpgtd %ymm6, %ymm4, %ymm4
+; AVX2-NEXT:    vextracti128 $1, %ymm4, %xmm6
+; AVX2-NEXT:    vpacksswb %xmm6, %xmm4, %xmm4
+; AVX2-NEXT:    vpshufb %xmm3, %xmm4, %xmm3
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
+; AVX2-NEXT:    vpsllw $7, %xmm3, %xmm3
+; AVX2-NEXT:    vpand %xmm1, %xmm3, %xmm1
+; AVX2-NEXT:    vpcmpgtb %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: v16i32:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vpcmpgtd %zmm1, %zmm0, %k1
+; AVX512F-NEXT:    vpcmpgtd %zmm3, %zmm2, %k0 {%k1}
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: v16i32:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpcmpgtd %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpcmpgtd %zmm3, %zmm2, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %x0 = icmp sgt <16 x i32> %a, %b
+  %x1 = icmp sgt <16 x i32> %c, %d
+  %y = and <16 x i1> %x0, %x1
+  %res = bitcast <16 x i1> %y to i16
+  ret i16 %res
+}
+
+define i16 @v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c, <16 x float> %d) {
+; SSE-LABEL: v16f32:
+; SSE:       # BB#0:
+; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    movaps {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    cmpltps %xmm3, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
+; SSE-NEXT:    pshufb %xmm3, %xmm7
+; SSE-NEXT:    cmpltps %xmm2, %xmm6
+; SSE-NEXT:    pshufb %xmm3, %xmm6
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm7[0]
+; SSE-NEXT:    psllw $15, %xmm6
+; SSE-NEXT:    psraw $15, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; SSE-NEXT:    pshufb %xmm2, %xmm6
+; SSE-NEXT:    cmpltps %xmm1, %xmm5
+; SSE-NEXT:    pshufb %xmm3, %xmm5
+; SSE-NEXT:    cmpltps %xmm0, %xmm4
+; SSE-NEXT:    pshufb %xmm3, %xmm4
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm5[0]
+; SSE-NEXT:    psllw $15, %xmm4
+; SSE-NEXT:    psraw $15, %xmm4
+; SSE-NEXT:    pshufb %xmm2, %xmm4
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    psllw $7, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    xorps %xmm0, %xmm0
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    pcmpgtb %xmm4, %xmm5
+; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pshufb %xmm3, %xmm11
+; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    pshufb %xmm3, %xmm9
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm11[0]
+; SSE-NEXT:    psllw $15, %xmm9
+; SSE-NEXT:    psraw $15, %xmm9
+; SSE-NEXT:    pshufb %xmm2, %xmm9
+; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    pshufb %xmm3, %xmm10
+; SSE-NEXT:    cmpltps {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    pshufb %xmm3, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm10[0]
+; SSE-NEXT:    psllw $15, %xmm8
+; SSE-NEXT:    psraw $15, %xmm8
+; SSE-NEXT:    pshufb %xmm2, %xmm8
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm9[0]
+; SSE-NEXT:    psllw $7, %xmm8
+; SSE-NEXT:    pand %xmm1, %xmm8
+; SSE-NEXT:    pcmpgtb %xmm8, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pmovmskb %xmm0, %eax
+; SSE-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; SSE-NEXT:    retq
+;
+; AVX12-LABEL: v16f32:
+; AVX12:       # BB#0:
+; AVX12-NEXT:    vcmpltps %ymm1, %ymm3, %ymm1
+; AVX12-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX12-NEXT:    vpacksswb %xmm3, %xmm1, %xmm1
+; AVX12-NEXT:    vmovdqa {{.*#+}} xmm3 = <0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u>
+; AVX12-NEXT:    vpshufb %xmm3, %xmm1, %xmm1
+; AVX12-NEXT:    vcmpltps %ymm0, %ymm2, %ymm0
+; AVX12-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX12-NEXT:    vpacksswb %xmm2, %xmm0, %xmm0
+; AVX12-NEXT:    vpshufb %xmm3, %xmm0, %xmm0
+; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX12-NEXT:    vpsllw $7, %xmm0, %xmm0
+; AVX12-NEXT:    vmovdqa {{.*#+}} xmm1 = [128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX12-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
+; AVX12-NEXT:    vcmpltps %ymm5, %ymm7, %ymm5
+; AVX12-NEXT:    vextractf128 $1, %ymm5, %xmm7
+; AVX12-NEXT:    vpacksswb %xmm7, %xmm5, %xmm5
+; AVX12-NEXT:    vpshufb %xmm3, %xmm5, %xmm5
+; AVX12-NEXT:    vcmpltps %ymm4, %ymm6, %ymm4
+; AVX12-NEXT:    vextractf128 $1, %ymm4, %xmm6
+; AVX12-NEXT:    vpacksswb %xmm6, %xmm4, %xmm4
+; AVX12-NEXT:    vpshufb %xmm3, %xmm4, %xmm3
+; AVX12-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm5[0]
+; AVX12-NEXT:    vpsllw $7, %xmm3, %xmm3
+; AVX12-NEXT:    vpand %xmm1, %xmm3, %xmm1
+; AVX12-NEXT:    vpcmpgtb %xmm1, %xmm2, %xmm1
+; AVX12-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX12-NEXT:    vpmovmskb %xmm0, %eax
+; AVX12-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX12-NEXT:    vzeroupper
+; AVX12-NEXT:    retq
+;
+; AVX512F-LABEL: v16f32:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    vcmpltps %zmm0, %zmm1, %k1
+; AVX512F-NEXT:    vcmpltps %zmm2, %zmm3, %k0 {%k1}
+; AVX512F-NEXT:    kmovw %k0, %eax
+; AVX512F-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: v16f32:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vcmpltps %zmm0, %zmm1, %k1
+; AVX512BW-NEXT:    vcmpltps %zmm2, %zmm3, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %x0 = fcmp ogt <16 x float> %a, %b
+  %x1 = fcmp ogt <16 x float> %c, %d
+  %y = and <16 x i1> %x0, %x1
+  %res = bitcast <16 x i1> %y to i16
+  ret i16 %res
+}
+
+define i64 @v64i8(<64 x i8> %a, <64 x i8> %b, <64 x i8> %c, <64 x i8> %d) {
+; SSE-LABEL: v64i8:
+; SSE:       # BB#0:
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    movdqa {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    pcmpgtb %xmm6, %xmm2
+; SSE-NEXT:    pcmpgtb %xmm7, %xmm3
+; SSE-NEXT:    pcmpgtb %xmm4, %xmm0
+; SSE-NEXT:    pcmpgtb %xmm5, %xmm1
+; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm8
+; SSE-NEXT:    pand %xmm2, %xmm8
+; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm9
+; SSE-NEXT:    pand %xmm3, %xmm9
+; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm10
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    pcmpgtb {{[0-9]+}}(%rsp), %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    pextrb $15, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $14, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $13, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $12, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $11, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $10, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $9, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $8, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $7, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $6, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $5, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $4, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $3, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $2, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $1, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $0, %xmm11, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $15, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $14, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $13, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $12, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $11, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $10, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $9, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $8, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $7, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $6, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $5, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $4, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $3, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $2, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $1, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $0, %xmm10, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $15, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $14, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $13, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $12, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $11, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $10, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $9, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $8, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $7, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $6, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $5, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $4, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $3, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $2, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $1, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $0, %xmm9, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $15, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $14, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $13, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $12, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $11, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $10, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $9, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $8, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $7, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $6, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $5, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $4, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $3, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $2, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $1, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    pextrb $0, %xmm8, %eax
+; SSE-NEXT:    andb $1, %al
+; SSE-NEXT:    movb %al, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movl -{{[0-9]+}}(%rsp), %eax
+; SSE-NEXT:    shll $16, %eax
+; SSE-NEXT:    movzwl -{{[0-9]+}}(%rsp), %ecx
+; SSE-NEXT:    orl %eax, %ecx
+; SSE-NEXT:    movl -{{[0-9]+}}(%rsp), %edx
+; SSE-NEXT:    shll $16, %edx
+; SSE-NEXT:    movzwl -{{[0-9]+}}(%rsp), %eax
+; SSE-NEXT:    orl %edx, %eax
+; SSE-NEXT:    shlq $32, %rax
+; SSE-NEXT:    orq %rcx, %rax
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: v64i8:
+; 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 $64, %rsp
+; AVX1-NEXT:    vextractf128 $1, %ymm3, %xmm8
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm9
+; AVX1-NEXT:    vpcmpgtb %xmm8, %xmm9, %xmm8
+; AVX1-NEXT:    vpcmpgtb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm8, %ymm1, %ymm8
+; AVX1-NEXT:    vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpgtb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpgtb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm7, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm5, %xmm2
+; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm2, %xmm0
+; AVX1-NEXT:    vpcmpgtb %xmm7, %xmm5, %xmm2
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT:    vandps %ymm0, %ymm8, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm6, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm4, %xmm3
+; AVX1-NEXT:    vpcmpgtb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpcmpgtb %xmm6, %xmm4, %xmm3
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpextrb $15, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $14, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $13, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $12, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $11, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $10, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $9, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $8, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $7, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $6, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $5, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $4, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $3, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $2, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $1, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; AVX1-NEXT:    vpextrb $0, %xmm2, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, (%rsp)
+; 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:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpextrb $15, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $14, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $13, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $12, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $11, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $10, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $9, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $8, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $7, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $6, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $5, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $4, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $3, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $2, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $1, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX1-NEXT:    vpextrb $0, %xmm1, %eax
+; AVX1-NEXT:    andb $1, %al
+; AVX1-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; 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:    movl (%rsp), %ecx
+; AVX1-NEXT:    movl {{[0-9]+}}(%rsp), %eax
+; AVX1-NEXT:    shlq $32, %rax
+; AVX1-NEXT:    orq %rcx, %rax
+; AVX1-NEXT:    movq %rbp, %rsp
+; AVX1-NEXT:    popq %rbp
+; AVX1-NEXT:    vzeroupper
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: v64i8:
+; 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 $64, %rsp
+; AVX2-NEXT:    vpcmpgtb %ymm3, %ymm1, %ymm1
+; AVX2-NEXT:    vpcmpgtb %ymm2, %ymm0, %ymm2
+; AVX2-NEXT:    vpcmpgtb %ymm7, %ymm5, %ymm0
+; AVX2-NEXT:    vpand %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpcmpgtb %ymm6, %ymm4, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpextrb $15, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $14, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $13, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $12, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $11, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $10, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $9, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $8, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $7, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $6, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $5, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $4, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $3, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $2, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $1, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; AVX2-NEXT:    vpextrb $0, %xmm2, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, (%rsp)
+; 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:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vpextrb $15, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $14, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $13, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $12, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $11, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $10, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $9, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $8, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $7, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $6, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $5, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $4, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $3, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $2, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $1, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; AVX2-NEXT:    vpextrb $0, %xmm1, %eax
+; AVX2-NEXT:    andb $1, %al
+; AVX2-NEXT:    movb %al, {{[0-9]+}}(%rsp)
+; 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:    movl (%rsp), %ecx
+; AVX2-NEXT:    movl {{[0-9]+}}(%rsp), %eax
+; AVX2-NEXT:    shlq $32, %rax
+; AVX2-NEXT:    orq %rcx, %rax
+; AVX2-NEXT:    movq %rbp, %rsp
+; AVX2-NEXT:    popq %rbp
+; AVX2-NEXT:    vzeroupper
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: v64i8:
+; AVX512F:       # BB#0:
+; AVX512F-NEXT:    pushq %rbp
+; AVX512F-NEXT:  .Lcfi3:
+; AVX512F-NEXT:    .cfi_def_cfa_offset 16
+; AVX512F-NEXT:  .Lcfi4:
+; AVX512F-NEXT:    .cfi_offset %rbp, -16
+; AVX512F-NEXT:    movq %rsp, %rbp
+; AVX512F-NEXT:  .Lcfi5:
+; AVX512F-NEXT:    .cfi_def_cfa_register %rbp
+; AVX512F-NEXT:    andq $-32, %rsp
+; AVX512F-NEXT:    subq $64, %rsp
+; AVX512F-NEXT:    vpcmpgtb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpcmpgtb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtb %ymm7, %ymm5, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpcmpgtb %ymm6, %ymm4, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX512F-NEXT:    vpmovsxbd %xmm2, %zmm2
+; AVX512F-NEXT:    vpslld $31, %zmm2, %zmm2
+; AVX512F-NEXT:    vptestmd %zmm2, %zmm2, %k0
+; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
+; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
+; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; AVX512F-NEXT:    kmovw %k0, (%rsp)
+; AVX512F-NEXT:    vextracti128 $1, %ymm1, %xmm0
+; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
+; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
+; AVX512F-NEXT:    vpmovsxbd %xmm1, %zmm0
+; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k0
+; AVX512F-NEXT:    kmovw %k0, {{[0-9]+}}(%rsp)
+; AVX512F-NEXT:    movl (%rsp), %ecx
+; AVX512F-NEXT:    movl {{[0-9]+}}(%rsp), %eax
+; AVX512F-NEXT:    shlq $32, %rax
+; AVX512F-NEXT:    orq %rcx, %rax
+; AVX512F-NEXT:    movq %rbp, %rsp
+; AVX512F-NEXT:    popq %rbp
+; AVX512F-NEXT:    vzeroupper
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: v64i8:
+; AVX512BW:       # BB#0:
+; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpcmpgtb %zmm3, %zmm2, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    vzeroupper
+; AVX512BW-NEXT:    retq
+  %x0 = icmp sgt <64 x i8> %a, %b
+  %x1 = icmp sgt <64 x i8> %c, %d
+  %y = and <64 x i1> %x0, %x1
+  %res = bitcast <64 x i1> %y to i64
+  ret i64 %res
+}

llvm/trunk/test/CodeGen/X86/vselect-packss.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2      | FileCheck %s --check-prefix=SSE --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2    | FileCheck %s --check-prefix=SSE --check-prefix=SSE42
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx       | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2      | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f   | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F
 
 ;
 ; General cases - packing of vector comparison to legal vector result types
 ;
 
 define <16 x i8> @vselect_packss_v16i16(<16 x i16> %a0, <16 x i16> %a1, <16 x i8> %a2, <16 x i8> %a3) {
-; SSE-LABEL: vselect_packss_v16i16:
-; SSE:       # BB#0:
-; SSE-NEXT:    pcmpeqw %xmm3, %xmm1
-; SSE-NEXT:    pcmpeqw %xmm2, %xmm0
-; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    retq
+; SSE2-LABEL: vselect_packss_v16i16:
+; SSE2:       # BB#0:
+; SSE2-NEXT:    pcmpeqw %xmm3, %xmm1
+; SSE2-NEXT:    pcmpeqw %xmm2, %xmm0
+; SSE2-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-NEXT:    pand %xmm0, %xmm4
+; SSE2-NEXT:    pandn %xmm5, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE42-LABEL: vselect_packss_v16i16:
+; SSE42:       # BB#0:
+; SSE42-NEXT:    pcmpeqw %xmm3, %xmm1
+; SSE42-NEXT:    pcmpeqw %xmm2, %xmm0
+; SSE42-NEXT:    packsswb %xmm1, %xmm0
+; SSE42-NEXT:    pblendvb %xmm0, %xmm4, %xmm5
+; SSE42-NEXT:    movdqa %xmm5, %xmm0
+; SSE42-NEXT:    retq
 ;
 ; AVX1-LABEL: vselect_packss_v16i16:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
 ; AVX1-NEXT:    vpcmpeqw %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm1
-; AVX1-NEXT:    vpandn %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: vselect_packss_v16i16:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm1
-; AVX2-NEXT:    vpandn %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: vselect_packss_v16i16:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vpmovsxwd %ymm0, %zmm0
 ; AVX512-NEXT:    vpmovdb %zmm0, %xmm0
; AVX512-NEXT:    retq
   ret <16 x i8> %6
 }
 
 ;
 ; PACKSS case
 ;
 
 define <16 x i8> @vselect_packss(<16 x i16> %a0, <16 x i16> %a1, <16 x i8> %a2, <16 x i8> %a3) {
-; SSE-LABEL: vselect_packss:
-; SSE:       # BB#0:
-; SSE-NEXT:    pcmpeqw %xmm3, %xmm1
-; SSE-NEXT:    pcmpeqw %xmm2, %xmm0
-; SSE-NEXT:    packsswb %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    retq
+; SSE2-LABEL: vselect_packss:
+; SSE2:       # BB#0:
+; SSE2-NEXT:    pcmpeqw %xmm3, %xmm1
+; SSE2-NEXT:    pcmpeqw %xmm2, %xmm0
+; SSE2-NEXT:    packsswb %xmm1, %xmm0
+; SSE2-NEXT:    pand %xmm0, %xmm4
+; SSE2-NEXT:    pandn %xmm5, %xmm0
+; SSE2-NEXT:    por %xmm4, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE42-LABEL: vselect_packss:
+; SSE42:       # BB#0:
+; SSE42-NEXT:    pcmpeqw %xmm3, %xmm1
+; SSE42-NEXT:    pcmpeqw %xmm2, %xmm0
+; SSE42-NEXT:    packsswb %xmm1, %xmm0
+; SSE42-NEXT:    pblendvb %xmm0, %xmm4, %xmm5
+; SSE42-NEXT:    movdqa %xmm5, %xmm0
+; SSE42-NEXT:    retq
 ;
 ; AVX1-LABEL: vselect_packss:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm4
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm5
 ; AVX1-NEXT:    vpcmpeqw %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vpacksswb %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm1
-; AVX1-NEXT:    vpandn %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: vselect_packss:
 ; AVX2:       # BB#0:
 ; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm1
-; AVX2-NEXT:    vpandn %xmm3, %xmm0, %xmm0
-; AVX2-NEXT:    vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: vselect_packss:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
 ; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm1
 ; AVX512-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX512-NEXT:    vpand %xmm2, %xmm0, %xmm1
-; AVX512-NEXT:    vpandn %xmm3, %xmm0, %xmm0
-; AVX512-NEXT:    vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT:    vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
 ; AVX512-NEXT:    vzeroupper
 ; AVX512-NEXT:    retq
   %1 = icmp eq <16 x i16> %a0, %a1
   %2 = sext <16 x i1> %1 to <16 x i16>
   %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %4 = shufflevector <16 x i16> %2, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   %5 = tail call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %3, <8 x i16> %4)
   %6 = and <16 x i8> %5, %a2
   %7 = xor <16 x i8> %5, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
   %8 = and <16 x i8> %7, %a3
   %9 = or <16 x i8> %6, %8
   ret <16 x i8> %9
 }
 declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>)