diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index db315cc7a26..693da1a531a 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -30627,6 +30627,20 @@ static SDValue combineBitcast(SDNode *N, SelectionDAG &DAG,
     }
   }
 
+  // Try to remove a bitcast of constant vXi1 vector. We have to legalize
+  // most of these to scalar anyway.
+  if (Subtarget.hasAVX512() && VT.isScalarInteger() &&
+      SrcVT.isVector() && SrcVT.getVectorElementType() == MVT::i1 &&
+      ISD::isBuildVectorOfConstantSDNodes(N0.getNode())) {
+    APInt Imm(SrcVT.getVectorNumElements(), 0);
+    for (unsigned Idx = 0, e = N0.getNumOperands(); Idx < e; ++Idx) {
+      SDValue In = N0.getOperand(Idx);
+      if (!In.isUndef() && (cast<ConstantSDNode>(In)->getZExtValue() & 0x1))
+        Imm.setBit(Idx);
+    }
+    return DAG.getConstant(Imm, SDLoc(N), VT);
+  }
+
   // Try to remove bitcasts from input and output of mask arithmetic to
   // remove GPR<->K-register crossings.
   if (SDValue V = combineCastedMaskArithmetic(N, DAG, DCI, Subtarget))
diff --git a/test/CodeGen/X86/avx512-intrinsics-upgrade.ll b/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
index 642c82728cf..f19f062f066 100644
--- a/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
+++ b/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
@@ -3164,9 +3164,8 @@ define <8 x i16> @test_cmp_d_512(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0
-; CHECK-NEXT:    kmovw %k0, %eax
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; CHECK-NEXT:    retq
   %res0 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 -1)
   %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
@@ -3194,21 +3193,19 @@ define <8 x i16> @test_mask_cmp_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask)
 ; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0 {%k1}
 ; CHECK-NEXT:    vpcmpgtd %zmm0, %zmm1, %k2 {%k1}
 ; CHECK-NEXT:    vpcmpled %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    kxorw %k0, %k0, %k4
-; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k6 {%k1}
+; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k4 {%k1}
+; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k5 {%k1}
 ; CHECK-NEXT:    vpcmpgtd %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
 ; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k3, %eax
 ; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
+; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k1, %eax
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
@@ -3257,9 +3254,8 @@ define <8 x i16> @test_ucmp_d_512(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0
-; CHECK-NEXT:    kmovw %k0, %eax
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; CHECK-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; CHECK-NEXT:    retq
   %res0 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 -1)
   %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
@@ -3287,21 +3283,19 @@ define <8 x i16> @test_mask_ucmp_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask
 ; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0 {%k1}
 ; CHECK-NEXT:    vpcmpltud %zmm1, %zmm0, %k2 {%k1}
 ; CHECK-NEXT:    vpcmpleud %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    kxorw %k0, %k0, %k4
-; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k6 {%k1}
+; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k4 {%k1}
+; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k5 {%k1}
 ; CHECK-NEXT:    vpcmpnleud %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
 ; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k3, %eax
 ; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
+; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k1, %eax
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
@@ -3350,8 +3344,7 @@ define <8 x i8> @test_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0
-; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    movl $255, %eax
 ; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
@@ -3380,25 +3373,23 @@ define <8 x i8> @test_mask_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
 ; CHECK-NEXT:    vpcmpgtq %zmm0, %zmm1, %k2 {%k1}
 ; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    kxorw %k0, %k0, %k4
-; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpleq %zmm0, %zmm1, %k6 {%k1}
+; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k4 {%k1}
+; CHECK-NEXT:    vpcmpleq %zmm0, %zmm1, %k5 {%k1}
 ; CHECK-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -3443,8 +3434,7 @@ define <8 x i8> @test_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k5, %eax
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0
-; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    movl $255, %eax
 ; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
@@ -3473,25 +3463,23 @@ define <8 x i8> @test_mask_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
 ; CHECK-NEXT:    vpcmpltuq %zmm1, %zmm0, %k2 {%k1}
 ; CHECK-NEXT:    vpcmpleuq %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    kxorw %k0, %k0, %k4
-; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k6 {%k1}
+; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k4 {%k1}
+; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k5 {%k1}
 ; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
diff --git a/test/CodeGen/X86/avx512-mask-op.ll b/test/CodeGen/X86/avx512-mask-op.ll
index 6953cd9ed47..f035b724f85 100644
--- a/test/CodeGen/X86/avx512-mask-op.ll
+++ b/test/CodeGen/X86/avx512-mask-op.ll
@@ -31,37 +31,11 @@ define i32 @mask16_zext(i16 %x) {
 }
 
 define i8 @mask8(i8 %x) {
-; KNL-LABEL: mask8:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kxnorw %k0, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    xorb %dil, %al
-; KNL-NEXT:    ## kill: def $al killed $al killed $eax
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: mask8:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    kxnorw %k0, %k0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    xorb %dil, %al
-; SKX-NEXT:    ## kill: def $al killed $al killed $eax
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: mask8:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kxnorw %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    xorb %dil, %al
-; AVX512BW-NEXT:    ## kill: def $al killed $al killed $eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: mask8:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    kxnorw %k0, %k0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    xorb %dil, %al
-; AVX512DQ-NEXT:    ## kill: def $al killed $al killed $eax
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: mask8:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    notb %dil
+; CHECK-NEXT:    movl %edi, %eax
+; CHECK-NEXT:    retq
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = xor <8 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1>
   %ret = bitcast <8 x i1> %m1 to i8
@@ -69,37 +43,11 @@ define i8 @mask8(i8 %x) {
 }
 
 define i32 @mask8_zext(i8 %x) {
-; KNL-LABEL: mask8_zext:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    kxnorw %k0, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    xorb %dil, %al
-; KNL-NEXT:    movzbl %al, %eax
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: mask8_zext:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    kxnorw %k0, %k0, %k0
-; SKX-NEXT:    kmovd %k0, %eax
-; SKX-NEXT:    xorb %dil, %al
-; SKX-NEXT:    movzbl %al, %eax
-; SKX-NEXT:    retq
-;
-; AVX512BW-LABEL: mask8_zext:
-; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kxnorw %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    xorb %dil, %al
-; AVX512BW-NEXT:    movzbl %al, %eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512DQ-LABEL: mask8_zext:
-; AVX512DQ:       ## %bb.0:
-; AVX512DQ-NEXT:    kxnorw %k0, %k0, %k0
-; AVX512DQ-NEXT:    kmovw %k0, %eax
-; AVX512DQ-NEXT:    xorb %dil, %al
-; AVX512DQ-NEXT:    movzbl %al, %eax
-; AVX512DQ-NEXT:    retq
+; CHECK-LABEL: mask8_zext:
+; CHECK:       ## %bb.0:
+; CHECK-NEXT:    notb %dil
+; CHECK-NEXT:    movzbl %dil, %eax
+; CHECK-NEXT:    retq
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = xor <8 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1>
   %m2 = bitcast <8 x i1> %m1 to i8
@@ -125,9 +73,7 @@ define void @mask16_mem(i16* %ptr) {
 define void @mask8_mem(i8* %ptr) {
 ; KNL-LABEL: mask8_mem:
 ; KNL:       ## %bb.0:
-; KNL-NEXT:    kxnorw %k0, %k0, %k0
-; KNL-NEXT:    kmovw %k0, %eax
-; KNL-NEXT:    xorb %al, (%rdi)
+; KNL-NEXT:    notb (%rdi)
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: mask8_mem:
@@ -139,9 +85,7 @@ define void @mask8_mem(i8* %ptr) {
 ;
 ; AVX512BW-LABEL: mask8_mem:
 ; AVX512BW:       ## %bb.0:
-; AVX512BW-NEXT:    kxnorw %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    xorb %al, (%rdi)
+; AVX512BW-NEXT:    notb (%rdi)
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: mask8_mem:
diff --git a/test/CodeGen/X86/avx512-schedule.ll b/test/CodeGen/X86/avx512-schedule.ll
index 1055570de07..99798c72d44 100755
--- a/test/CodeGen/X86/avx512-schedule.ll
+++ b/test/CodeGen/X86/avx512-schedule.ll
@@ -6700,18 +6700,14 @@ define i32 @mask16_zext(i16 %x) {
 define i8 @mask8(i8 %x) {
 ; GENERIC-LABEL: mask8:
 ; GENERIC:       # %bb.0:
-; GENERIC-NEXT:    kxnorw %k0, %k0, %k0 # sched: [1:1.00]
-; GENERIC-NEXT:    kmovd %k0, %eax # sched: [1:0.33]
-; GENERIC-NEXT:    xorb %dil, %al # sched: [1:0.33]
-; GENERIC-NEXT:    # kill: def $al killed $al killed $eax
+; GENERIC-NEXT:    notb %dil # sched: [1:0.33]
+; GENERIC-NEXT:    movl %edi, %eax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: mask8:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    kxnorw %k0, %k0, %k0 # sched: [1:1.00]
-; SKX-NEXT:    kmovd %k0, %eax # sched: [3:1.00]
-; SKX-NEXT:    xorb %dil, %al # sched: [1:0.25]
-; SKX-NEXT:    # kill: def $al killed $al killed $eax
+; SKX-NEXT:    notb %dil # sched: [1:0.25]
+; SKX-NEXT:    movl %edi, %eax # sched: [1:0.25]
 ; SKX-NEXT:    retq # sched: [7:1.00]
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = xor <8 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1>
@@ -6722,18 +6718,14 @@ define i8 @mask8(i8 %x) {
 define i32 @mask8_zext(i8 %x) {
 ; GENERIC-LABEL: mask8_zext:
 ; GENERIC:       # %bb.0:
-; GENERIC-NEXT:    kxnorw %k0, %k0, %k0 # sched: [1:1.00]
-; GENERIC-NEXT:    kmovd %k0, %eax # sched: [1:0.33]
-; GENERIC-NEXT:    xorb %dil, %al # sched: [1:0.33]
-; GENERIC-NEXT:    movzbl %al, %eax # sched: [1:0.33]
+; GENERIC-NEXT:    notb %dil # sched: [1:0.33]
+; GENERIC-NEXT:    movzbl %dil, %eax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: mask8_zext:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    kxnorw %k0, %k0, %k0 # sched: [1:1.00]
-; SKX-NEXT:    kmovd %k0, %eax # sched: [3:1.00]
-; SKX-NEXT:    xorb %dil, %al # sched: [1:0.25]
-; SKX-NEXT:    movzbl %al, %eax # sched: [1:0.25]
+; SKX-NEXT:    notb %dil # sched: [1:0.25]
+; SKX-NEXT:    movzbl %dil, %eax # sched: [1:0.25]
 ; SKX-NEXT:    retq # sched: [7:1.00]
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = xor <8 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1>
diff --git a/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll b/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
index 282079e1b52..b00fac6fbf3 100644
--- a/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
+++ b/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
@@ -1693,10 +1693,7 @@ define i64 @test_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
 ; AVX512BW-NEXT:    addq %rcx, %rax
 ; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    kxnorq %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    leaq -1(%rcx,%rax), %rax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
@@ -1742,10 +1739,8 @@ define i64 @test_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
 ; AVX512F-32-NEXT:    kmovd %k0, %eax
 ; AVX512F-32-NEXT:    addl %edi, %eax
 ; AVX512F-32-NEXT:    adcl %ecx, %edx
-; AVX512F-32-NEXT:    kxnord %k0, %k0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    adcl %ecx, %edx
+; AVX512F-32-NEXT:    addl $-1, %eax
+; AVX512F-32-NEXT:    adcl $-1, %edx
 ; AVX512F-32-NEXT:    popl %esi
 ; AVX512F-32-NEXT:    popl %edi
 ; AVX512F-32-NEXT:    vzeroupper
@@ -1904,10 +1899,7 @@ define i64 @test_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
 ; AVX512BW-NEXT:    addq %rcx, %rax
 ; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    kxnorq %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    leaq -1(%rcx,%rax), %rax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
@@ -1953,10 +1945,8 @@ define i64 @test_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
 ; AVX512F-32-NEXT:    kmovd %k0, %eax
 ; AVX512F-32-NEXT:    addl %edi, %eax
 ; AVX512F-32-NEXT:    adcl %ecx, %edx
-; AVX512F-32-NEXT:    kxnord %k0, %k0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    adcl %ecx, %edx
+; AVX512F-32-NEXT:    addl $-1, %eax
+; AVX512F-32-NEXT:    adcl $-1, %edx
 ; AVX512F-32-NEXT:    popl %esi
 ; AVX512F-32-NEXT:    popl %edi
 ; AVX512F-32-NEXT:    vzeroupper
@@ -2115,10 +2105,7 @@ define i32 @test_cmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
 ; AVX512BW-NEXT:    addl %ecx, %eax
 ; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    kxnord %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    leal -1(%rcx,%rax), %eax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
@@ -2140,10 +2127,7 @@ define i32 @test_cmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
 ; AVX512F-32-NEXT:    addl %ecx, %eax
 ; AVX512F-32-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0
 ; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    kxnord %k0, %k0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    leal -1(%ecx,%eax), %eax
 ; AVX512F-32-NEXT:    vzeroupper
 ; AVX512F-32-NEXT:    retl
   %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
@@ -2256,10 +2240,7 @@ define i32 @test_ucmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
 ; AVX512BW-NEXT:    addl %ecx, %eax
 ; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    kxnord %k0, %k0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    leal -1(%rcx,%rax), %eax
 ; AVX512BW-NEXT:    vzeroupper
 ; AVX512BW-NEXT:    retq
 ;
@@ -2281,10 +2262,7 @@ define i32 @test_ucmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
 ; AVX512F-32-NEXT:    addl %ecx, %eax
 ; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
 ; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    kxnord %k0, %k0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    leal -1(%ecx,%eax), %eax
 ; AVX512F-32-NEXT:    vzeroupper
 ; AVX512F-32-NEXT:    retl
   %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
diff --git a/test/CodeGen/X86/avx512bw-mask-op.ll b/test/CodeGen/X86/avx512bw-mask-op.ll
index 7fc421c0ece..7ddb1f34d6b 100644
--- a/test/CodeGen/X86/avx512bw-mask-op.ll
+++ b/test/CodeGen/X86/avx512bw-mask-op.ll
@@ -4,9 +4,8 @@
 define i32 @mask32(i32 %x) {
 ; CHECK-LABEL: mask32:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kxnord %k0, %k0, %k0
-; CHECK-NEXT:    kmovd %k0, %eax
-; CHECK-NEXT:    xorl %edi, %eax
+; CHECK-NEXT:    notl %edi
+; CHECK-NEXT:    movl %edi, %eax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i32 %x to <32 x i1>
   %m1 = xor <32 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1,
@@ -20,9 +19,8 @@ define i32 @mask32(i32 %x) {
 define i64 @mask64(i64 %x) {
 ; CHECK-LABEL: mask64:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kxnorq %k0, %k0, %k0
-; CHECK-NEXT:    kmovq %k0, %rax
-; CHECK-NEXT:    xorq %rdi, %rax
+; CHECK-NEXT:    notq %rdi
+; CHECK-NEXT:    movq %rdi, %rax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i64 %x to <64 x i1>
   %m1 = xor <64 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1,
diff --git a/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll b/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll
index ab52a7976b7..416d5a7bdeb 100644
--- a/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll
+++ b/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll
@@ -2804,9 +2804,9 @@ define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
 ; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
 ; CHECK-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
 ; CHECK-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
-; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x76,0xc9]
+; CHECK-NEXT:    vpblendd $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
+; CHECK-NEXT:    ## xmm0 = xmm0[0,1,2],xmm1[3]
 ; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
 ; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
 ; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
@@ -2842,25 +2842,26 @@ define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
 ; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
 ; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
 ; CHECK-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
-; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
-; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x3f,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
+; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x3f,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
 ; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
 ; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01]
+; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
 ; CHECK-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
 ; CHECK-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
-; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
-; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
+; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
+; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
+; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
+; CHECK-NEXT:    ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-NEXT:    vmovd %r9d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
+; CHECK-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
+; CHECK-NEXT:    ## xmm1 = xmm1[0],xmm2[0]
 ; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
@@ -2902,9 +2903,9 @@ define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
 ; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
 ; CHECK-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
 ; CHECK-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
-; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x76,0xc9]
+; CHECK-NEXT:    vpblendd $8, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
+; CHECK-NEXT:    ## xmm0 = xmm0[0,1,2],xmm1[3]
 ; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
 ; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
 ; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
@@ -2940,25 +2941,26 @@ define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask)
 ; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
 ; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
 ; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
-; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
-; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
 ; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
+; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
 ; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
 ; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01]
+; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
 ; CHECK-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
 ; CHECK-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
-; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
-; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
+; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
+; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
+; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
+; CHECK-NEXT:    ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-NEXT:    vmovd %r9d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
+; CHECK-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
+; CHECK-NEXT:    ## xmm1 = xmm1[0],xmm2[0]
 ; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
@@ -3004,9 +3006,9 @@ define <8 x i16> @test_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x76,0xc9]
+; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
+; CHECK-NEXT:    ## xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; CHECK-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
@@ -3035,21 +3037,19 @@ define <8 x i16> @test_mask_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask)
 ; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
 ; CHECK-NEXT:    vpcmpgtw %ymm0, %ymm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
 ; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmplew %ymm0, %ymm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x3f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
+; CHECK-NEXT:    vpcmplew %ymm0, %ymm1, %k5 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x3f,0xe8,0x02]
 ; CHECK-NEXT:    vpcmpgtw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
 ; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
@@ -3099,9 +3099,9 @@ define <8 x i16> @test_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x76,0xc9]
+; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
+; CHECK-NEXT:    ## xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; CHECK-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
@@ -3130,21 +3130,19 @@ define <8 x i16> @test_mask_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask
 ; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
 ; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
 ; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
 ; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
 ; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
@@ -3194,9 +3192,9 @@ define <8 x i16> @test_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x76,0xc9]
+; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
+; CHECK-NEXT:    ## xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
   %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
@@ -3224,21 +3222,19 @@ define <8 x i16> @test_mask_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
 ; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
 ; CHECK-NEXT:    vpcmpgtb %xmm0, %xmm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
 ; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpleb %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0x75,0x09,0x3f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpleb %xmm0, %xmm1, %k5 {%k1} ## encoding: [0x62,0xf3,0x75,0x09,0x3f,0xe8,0x02]
 ; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
 ; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
@@ -3287,9 +3283,9 @@ define <8 x i16> @test_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x76,0xc9]
+; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
+; CHECK-NEXT:    ## xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
   %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
@@ -3317,21 +3313,19 @@ define <8 x i16> @test_mask_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask)
 ; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
 ; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
 ; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
 ; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
 ; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
@@ -3380,8 +3374,7 @@ define <8 x i8> @test_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
 ; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
@@ -3410,25 +3403,23 @@ define <8 x i8> @test_mask_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
 ; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
 ; CHECK-NEXT:    vpcmpgtw %xmm0, %xmm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
 ; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmplew %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x09,0x3f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
+; CHECK-NEXT:    vpcmplew %xmm0, %xmm1, %k5 {%k1} ## encoding: [0x62,0xf3,0xf5,0x09,0x3f,0xe8,0x02]
 ; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -3473,8 +3464,7 @@ define <8 x i8> @test_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
 ; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
@@ -3503,25 +3493,23 @@ define <8 x i8> @test_mask_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
 ; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
 ; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
 ; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
 ; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
 ; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
diff --git a/test/CodeGen/X86/avx512dq-mask-op.ll b/test/CodeGen/X86/avx512dq-mask-op.ll
index 96b45fd91c8..7bc367f882d 100644
--- a/test/CodeGen/X86/avx512dq-mask-op.ll
+++ b/test/CodeGen/X86/avx512dq-mask-op.ll
@@ -4,10 +4,8 @@
 define i8 @mask8(i8 %x) {
 ; CHECK-LABEL: mask8:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0
-; CHECK-NEXT:    kmovd %k0, %eax
-; CHECK-NEXT:    xorb %dil, %al
-; CHECK-NEXT:    ## kill: def $al killed $al killed $eax
+; CHECK-NEXT:    notb %dil
+; CHECK-NEXT:    movl %edi, %eax
 ; CHECK-NEXT:    retq
   %m0 = bitcast i8 %x to <8 x i1>
   %m1 = xor <8 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1>
diff --git a/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll b/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
index 589d9a4dbb7..f56d5b34e48 100644
--- a/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
+++ b/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
@@ -5124,8 +5124,7 @@ define <8 x i8> @test_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
 ; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
@@ -5154,25 +5153,23 @@ define <8 x i8> @test_mask_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
 ; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
 ; CHECK-NEXT:    vpcmpgtd %ymm0, %ymm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x66,0xd0]
 ; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpled %ymm0, %ymm1, %k6 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x1f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpled %ymm0, %ymm1, %k5 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x1f,0xe8,0x02]
 ; CHECK-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc9]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
 ; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5217,8 +5214,7 @@ define <8 x i8> @test_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
 ; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
 ; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
 ; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
@@ -5247,25 +5243,23 @@ define <8 x i8> @test_mask_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
 ; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
 ; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x01]
 ; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd9,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe9,0x05]
 ; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x06]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
 ; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
 ; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
 ; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5335,32 +5329,30 @@ define <8 x i8> @test_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
 define <8 x i8> @test_mask_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: test_mask_cmp_q_256:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
-; CHECK-NEXT:    vpcmpgtq %ymm0, %ymm1, %k3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0x37,0xd8]
-; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xd1,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpleq %ymm0, %ymm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x1f,0xf0,0x02]
-; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x37,0xf9]
-; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
-; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %edi, %k2 ## encoding: [0xc5,0xf8,0x92,0xd7]
+; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k2} ## encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
+; CHECK-NEXT:    vpcmpgtq %ymm0, %ymm1, %k1 {%k2} ## encoding: [0x62,0xf2,0xf5,0x2a,0x37,0xc8]
+; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k3 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xd9,0x02]
+; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpleq %ymm0, %ymm1, %k5 {%k2} ## encoding: [0x62,0xf3,0xf5,0x2a,0x1f,0xe8,0x02]
+; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k6 {%k2} ## encoding: [0x62,0xf2,0xfd,0x2a,0x37,0xf1]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5430,32 +5422,30 @@ define <8 x i8> @test_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
 define <8 x i8> @test_mask_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: test_mask_ucmp_q_256:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
-; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd9,0x01]
-; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd1,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf9,0x06]
-; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
-; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %edi, %k2 ## encoding: [0xc5,0xf8,0x92,0xd7]
+; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k2} ## encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
+; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k1 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k3 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xd9,0x02]
+; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k5 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xe9,0x05]
+; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k6 {%k2} ## encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xf1,0x06]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5525,32 +5515,30 @@ define <8 x i8> @test_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
 define <8 x i8> @test_mask_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
 ; CHECK-LABEL: test_mask_cmp_d_128:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
-; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %k3 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x66,0xd8]
-; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xd1,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpled %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0x75,0x09,0x1f,0xf0,0x02]
-; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x66,0xf9]
-; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
-; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %edi, %k2 ## encoding: [0xc5,0xf8,0x92,0xd7]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k2} ## encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
+; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %k1 {%k2} ## encoding: [0x62,0xf1,0x75,0x0a,0x66,0xc8]
+; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k3 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xd9,0x02]
+; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpled %xmm0, %xmm1, %k5 {%k2} ## encoding: [0x62,0xf3,0x75,0x0a,0x1f,0xe8,0x02]
+; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %k6 {%k2} ## encoding: [0x62,0xf1,0x7d,0x0a,0x66,0xf1]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5620,32 +5608,30 @@ define <8 x i8> @test_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
 define <8 x i8> @test_mask_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
 ; CHECK-LABEL: test_mask_ucmp_d_128:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
-; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd9,0x01]
-; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd1,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf9,0x06]
-; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
-; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %edi, %k2 ## encoding: [0xc5,0xf8,0x92,0xd7]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k2} ## encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
+; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k1 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k3 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xd9,0x02]
+; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k5 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xe9,0x05]
+; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k6 {%k2} ## encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xf1,0x06]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5715,32 +5701,30 @@ define <8 x i8> @test_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
 define <8 x i8> @test_mask_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: test_mask_cmp_q_128:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
-; CHECK-NEXT:    vpcmpgtq %xmm0, %xmm1, %k3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0x37,0xd8]
-; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xd1,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpleq %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x09,0x1f,0xf0,0x02]
-; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x37,0xf9]
-; CHECK-NEXT:    kshiftlw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
-; CHECK-NEXT:    kshiftrw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %edi, %k2 ## encoding: [0xc5,0xf8,0x92,0xd7]
+; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k2} ## encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
+; CHECK-NEXT:    vpcmpgtq %xmm0, %xmm1, %k1 {%k2} ## encoding: [0x62,0xf2,0xf5,0x0a,0x37,0xc8]
+; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k3 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xd9,0x02]
+; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpleq %xmm0, %xmm1, %k5 {%k2} ## encoding: [0x62,0xf3,0xf5,0x0a,0x1f,0xe8,0x02]
+; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k6 {%k2} ## encoding: [0x62,0xf2,0xfd,0x0a,0x37,0xf1]
+; CHECK-NEXT:    kshiftlw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
@@ -5810,32 +5794,30 @@ define <8 x i8> @test_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
 define <8 x i8> @test_mask_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: test_mask_ucmp_q_128:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
-; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd9,0x01]
-; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd1,0x02]
-; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
-; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf9,0x06]
-; CHECK-NEXT:    kshiftlw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
-; CHECK-NEXT:    kshiftrw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %edi, %k2 ## encoding: [0xc5,0xf8,0x92,0xd7]
+; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k2} ## encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
+; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k1 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k3 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xd9,0x02]
+; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
+; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k5 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xe9,0x05]
+; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k6 {%k2} ## encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xf1,0x06]
+; CHECK-NEXT:    kshiftlw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
 ; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0