diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -4244,6 +4244,9 @@
 } // ExeDomain = SSEPackedInt, isCodeGenOnly = 1
 
 let Predicates = [UseAVX] in {
+  def : Pat<(v4i32 (scalar_to_vector (i32 (anyext GR8:$src)))),
+            (VMOVDI2PDIrr (i32 (INSERT_SUBREG (i32 (IMPLICIT_DEF)),
+                                              GR8:$src, sub_8bit)))>;
   def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector GR32:$src)))),
             (VMOVDI2PDIrr GR32:$src)>;
 
@@ -5353,8 +5356,13 @@
                    Sched<[WriteVecInsert.Folded, WriteVecInsert.ReadAfterFold]>;
 }
 
-let Predicates = [HasAVX, NoBWI] in
+let Predicates = [HasAVX, NoBWI] in {
   defm VPINSRB : SS41I_insert8<0x20, "vpinsrb", 0>, VEX_4V, VEX_WIG;
+  def : Pat<(X86pinsrb VR128:$src1, (i32 (anyext (i8 GR8:$src2))), timm:$src3),
+            (VPINSRBrr VR128:$src1, (INSERT_SUBREG (i32 (IMPLICIT_DEF)),
+                       GR8:$src2, sub_8bit), timm:$src3)>;
+}
+
 let Constraints = "$src1 = $dst" in
   defm PINSRB  : SS41I_insert8<0x20, "pinsrb">;
 
diff --git a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx-intrinsics-fast-isel.ll
@@ -718,8 +718,7 @@
 ;
 ; X64-LABEL: test_mm256_insert_epi8:
 ; X64:       # %bb.0:
-; X64-NEXT:    movzbl %dil, %eax
-; X64-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm1
+; X64-NEXT:    vpinsrb $4, %edi, %xmm0, %xmm1
 ; X64-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <32 x i8>
@@ -1421,8 +1420,8 @@
 ; X86-LABEL: test_mm256_set_epi8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    vmovd %ecx, %xmm0
+; X86-NEXT:    vmovd %eax, %xmm0
+; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
@@ -1453,8 +1452,8 @@
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    vmovd %ecx, %xmm1
+; X86-NEXT:    vmovd %eax, %xmm1
+; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
@@ -1489,10 +1488,10 @@
 ;
 ; X64-LABEL: test_mm256_set_epi8:
 ; X64:       # %bb.0:
-; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %r10d
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vmovd %eax, %xmm0
-; X64-NEXT:    vpinsrb $1, %r10d, %xmm0, %xmm0
+; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
+; X64-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
@@ -1509,21 +1508,15 @@
 ; X64-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %r9b, %eax
-; X64-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %r8b, %eax
-; X64-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %cl, %eax
-; X64-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %dl, %eax
-; X64-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %sil, %eax
-; X64-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %dil, %eax
-; X64-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
+; X64-NEXT:    vpinsrb $10, %r9d, %xmm0, %xmm0
+; X64-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0
+; X64-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0
+; X64-NEXT:    vpinsrb $13, %edx, %xmm0, %xmm0
+; X64-NEXT:    vpinsrb $14, %esi, %xmm0, %xmm0
+; X64-NEXT:    vpinsrb $15, %edi, %xmm0, %xmm0
+; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
+; X64-NEXT:    vmovd %eax, %xmm1
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
-; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %ecx
-; X64-NEXT:    vmovd %ecx, %xmm1
 ; X64-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
@@ -1862,8 +1855,7 @@
 ;
 ; X64-LABEL: test_mm256_set1_epi8:
 ; X64:       # %bb.0:
-; X64-NEXT:    movzbl %dil, %eax
-; X64-NEXT:    vmovd %eax, %xmm0
+; X64-NEXT:    vmovd %edi, %xmm0
 ; X64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; X64-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
@@ -2037,8 +2029,8 @@
 ; X86-LABEL: test_mm256_setr_epi8:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    vmovd %ecx, %xmm0
+; X86-NEXT:    vmovd %eax, %xmm0
+; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
@@ -2069,8 +2061,8 @@
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    vmovd %ecx, %xmm1
+; X86-NEXT:    vmovd %eax, %xmm1
+; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
 ; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
@@ -2105,10 +2097,10 @@
 ;
 ; X64-LABEL: test_mm256_setr_epi8:
 ; X64:       # %bb.0:
-; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %r10d
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vmovd %eax, %xmm0
-; X64-NEXT:    vpinsrb $1, %r10d, %xmm0, %xmm0
+; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
+; X64-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
@@ -2137,18 +2129,12 @@
 ; X64-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
-; X64-NEXT:    movzbl %sil, %eax
-; X64-NEXT:    movzbl %dil, %esi
-; X64-NEXT:    vmovd %esi, %xmm1
-; X64-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
-; X64-NEXT:    movzbl %dl, %eax
-; X64-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
-; X64-NEXT:    movzbl %cl, %eax
-; X64-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
-; X64-NEXT:    movzbl %r8b, %eax
-; X64-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
-; X64-NEXT:    movzbl %r9b, %eax
-; X64-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
+; X64-NEXT:    vmovd %edi, %xmm1
+; X64-NEXT:    vpinsrb $1, %esi, %xmm1, %xmm1
+; X64-NEXT:    vpinsrb $2, %edx, %xmm1, %xmm1
+; X64-NEXT:    vpinsrb $3, %ecx, %xmm1, %xmm1
+; X64-NEXT:    vpinsrb $4, %r8d, %xmm1, %xmm1
+; X64-NEXT:    vpinsrb $5, %r9d, %xmm1, %xmm1
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
 ; X64-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
 ; X64-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
--- a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll
@@ -6209,31 +6209,69 @@
 declare i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32>, <16 x i32>, i32, i16) nounwind readnone
 
 define <8 x i8> @test_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
-; CHECK-LABEL: test_cmp_q_512:
-; CHECK:       ## %bb.0:
-; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x29,0xc1]
-; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
-; CHECK-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0 ## encoding: [0x62,0xf2,0xf5,0x48,0x37,0xc0]
-; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc9,0x02]
-; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xd1,0x04]
-; CHECK-NEXT:    vpcmpnltq %zmm1, %zmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xd9,0x05]
-; CHECK-NEXT:    vpcmpgtq %zmm1, %zmm0, %k4 ## encoding: [0x62,0xf2,0xfd,0x48,0x37,0xe1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    movzbl %al, %eax ## encoding: [0x0f,0xb6,0xc0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; CHECK-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; 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 %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; 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:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
-; CHECK-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
-; CHECK-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
-; CHECK-NEXT:    ret{{[l|q]}} ## encoding: [0xc3]
+; X86-LABEL: test_cmp_q_512:
+; X86:       ## %bb.0:
+; X86-NEXT:    pushl %ebx ## encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %edi ## encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %esi ## encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    .cfi_offset %esi, -16
+; X86-NEXT:    .cfi_offset %edi, -12
+; X86-NEXT:    .cfi_offset %ebx, -8
+; X86-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x29,0xc1]
+; X86-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X86-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0 ## encoding: [0x62,0xf2,0xf5,0x48,0x37,0xc0]
+; X86-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; X86-NEXT:    vpcmpleq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %edx ## encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpnltq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %edi ## encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpgtq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x37,0xc1]
+; X86-NEXT:    kmovw %k0, %ebx ## encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    movzbl %al, %eax ## encoding: [0x0f,0xb6,0xc0]
+; X86-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; X86-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
+; X86-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X86-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X86-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X86-NEXT:    vpinsrb $6, %ebx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x06]
+; X86-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi ## encoding: [0x5e]
+; X86-NEXT:    popl %edi ## encoding: [0x5f]
+; X86-NEXT:    popl %ebx ## encoding: [0x5b]
+; X86-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
+; X86-NEXT:    retl ## encoding: [0xc3]
+;
+; X64-LABEL: test_cmp_q_512:
+; X64:       ## %bb.0:
+; X64-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x29,0xc1]
+; X64-NEXT:    kmovw %k0, %r8d ## encoding: [0xc5,0x78,0x93,0xc0]
+; X64-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0 ## encoding: [0x62,0xf2,0xf5,0x48,0x37,0xc0]
+; X64-NEXT:    kmovw %k0, %r9d ## encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpleq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %edx ## encoding: [0xc5,0xf8,0x93,0xd0]
+; X64-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %edi ## encoding: [0xc5,0xf8,0x93,0xf8]
+; X64-NEXT:    vpcmpgtq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x37,0xc1]
+; X64-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    movzbl %r8b, %ecx ## encoding: [0x41,0x0f,0xb6,0xc8]
+; X64-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 ## encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X64-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X64-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X64-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
+; X64-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
   %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 -1)
@@ -6256,28 +6294,44 @@
 define <8 x i8> @test_mask_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; X86-LABEL: test_mask_cmp_q_512:
 ; X86:       ## %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
+; X86-NEXT:    pushl %ebp ## encoding: [0x55]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %ebx ## encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %edi ## encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    pushl %esi ## encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 20
+; X86-NEXT:    .cfi_offset %esi, -20
+; X86-NEXT:    .cfi_offset %edi, -16
+; X86-NEXT:    .cfi_offset %ebx, -12
+; X86-NEXT:    .cfi_offset %ebp, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x14]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x29,0xc1]
 ; X86-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
 ; X86-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x49,0x37,0xc0]
-; X86-NEXT:    vpcmpleq %zmm1, %zmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xd1,0x02]
-; X86-NEXT:    vpcmpneqq %zmm1, %zmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xd9,0x04]
-; X86-NEXT:    vpcmpnltq %zmm1, %zmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xe1,0x05]
-; X86-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x37,0xc9]
 ; X86-NEXT:    kmovw %k0, %edx ## encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpleq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %edi ## encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpnltq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %ebx ## encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    vpcmpgtq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x37,0xc1]
+; X86-NEXT:    kmovw %k0, %ebp ## encoding: [0xc5,0xf8,0x93,0xe8]
 ; X86-NEXT:    movzbl %cl, %ecx ## encoding: [0x0f,0xb6,0xc9]
 ; X86-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; X86-NEXT:    vpinsrb $1, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x01]
-; X86-NEXT:    kmovw %k2, %ecx ## encoding: [0xc5,0xf8,0x93,0xca]
-; X86-NEXT:    vpinsrb $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x02]
-; X86-NEXT:    kmovw %k3, %ecx ## encoding: [0xc5,0xf8,0x93,0xcb]
-; X86-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x04]
-; X86-NEXT:    kmovw %k4, %ecx ## encoding: [0xc5,0xf8,0x93,0xcc]
-; X86-NEXT:    vpinsrb $5, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x05]
-; X86-NEXT:    kmovw %k1, %ecx ## encoding: [0xc5,0xf8,0x93,0xc9]
-; X86-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
+; X86-NEXT:    vpinsrb $2, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x02]
+; X86-NEXT:    vpinsrb $4, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x04]
+; X86-NEXT:    vpinsrb $5, %ebx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x05]
+; X86-NEXT:    vpinsrb $6, %ebp, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc5,0x06]
 ; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi ## encoding: [0x5e]
+; X86-NEXT:    popl %edi ## encoding: [0x5f]
+; X86-NEXT:    popl %ebx ## encoding: [0x5b]
+; X86-NEXT:    popl %ebp ## encoding: [0x5d]
 ; X86-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
@@ -6285,24 +6339,24 @@
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x29,0xc1]
-; X64-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    kmovw %k0, %r8d ## encoding: [0xc5,0x78,0x93,0xc0]
 ; X64-NEXT:    vpcmpgtq %zmm0, %zmm1, %k0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x49,0x37,0xc0]
-; X64-NEXT:    vpcmpleq %zmm1, %zmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xd1,0x02]
-; X64-NEXT:    vpcmpneqq %zmm1, %zmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xd9,0x04]
-; X64-NEXT:    vpcmpnltq %zmm1, %zmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xe1,0x05]
-; X64-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x37,0xc9]
+; X64-NEXT:    kmovw %k0, %r9d ## encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpleq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %r10d ## encoding: [0xc5,0x78,0x93,0xd0]
+; X64-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    vpcmpgtq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x37,0xc1]
 ; X64-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; X64-NEXT:    movzbl %al, %eax ## encoding: [0x0f,0xb6,0xc0]
-; X64-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; X64-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; X64-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; X64-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; X64-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; X64-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; X64-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; X64-NEXT:    movzbl %r8b, %edx ## encoding: [0x41,0x0f,0xb6,0xd0]
+; X64-NEXT:    vmovd %edx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc2]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 ## encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %r10d, %xmm0, %xmm0 ## encoding: [0xc4,0xc3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
 ; X64-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; X64-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
 ; X64-NEXT:    vpinsrb $7, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x07]
 ; X64-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
 ; X64-NEXT:    retq ## encoding: [0xc3]
@@ -6328,31 +6382,69 @@
 declare i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64>, <8 x i64>, i32, i8) nounwind readnone
 
 define <8 x i8> @test_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
-; CHECK-LABEL: test_ucmp_q_512:
-; CHECK:       ## %bb.0:
-; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x29,0xc1]
-; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
-; CHECK-NEXT:    vpcmpltuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x01]
-; CHECK-NEXT:    vpcmpleuq %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc9,0x02]
-; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xd1,0x04]
-; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xd9,0x05]
-; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xe1,0x06]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    movzbl %al, %eax ## encoding: [0x0f,0xb6,0xc0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; CHECK-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; 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 %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; 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:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
-; CHECK-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
-; CHECK-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
-; CHECK-NEXT:    ret{{[l|q]}} ## encoding: [0xc3]
+; X86-LABEL: test_ucmp_q_512:
+; X86:       ## %bb.0:
+; X86-NEXT:    pushl %ebx ## encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %edi ## encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %esi ## encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    .cfi_offset %esi, -16
+; X86-NEXT:    .cfi_offset %edi, -12
+; X86-NEXT:    .cfi_offset %ebx, -8
+; X86-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x29,0xc1]
+; X86-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X86-NEXT:    vpcmpltuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x01]
+; X86-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; X86-NEXT:    vpcmpleuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %edx ## encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %edi ## encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x06]
+; X86-NEXT:    kmovw %k0, %ebx ## encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    movzbl %al, %eax ## encoding: [0x0f,0xb6,0xc0]
+; X86-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; X86-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
+; X86-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X86-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X86-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X86-NEXT:    vpinsrb $6, %ebx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x06]
+; X86-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi ## encoding: [0x5e]
+; X86-NEXT:    popl %edi ## encoding: [0x5f]
+; X86-NEXT:    popl %ebx ## encoding: [0x5b]
+; X86-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
+; X86-NEXT:    retl ## encoding: [0xc3]
+;
+; X64-LABEL: test_ucmp_q_512:
+; X64:       ## %bb.0:
+; X64-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x48,0x29,0xc1]
+; X64-NEXT:    kmovw %k0, %r8d ## encoding: [0xc5,0x78,0x93,0xc0]
+; X64-NEXT:    vpcmpltuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x01]
+; X64-NEXT:    kmovw %k0, %r9d ## encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpleuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %edx ## encoding: [0xc5,0xf8,0x93,0xd0]
+; X64-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %edi ## encoding: [0xc5,0xf8,0x93,0xf8]
+; X64-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x48,0x1e,0xc1,0x06]
+; X64-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    movzbl %r8b, %ecx ## encoding: [0x41,0x0f,0xb6,0xc8]
+; X64-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 ## encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X64-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    movl $255, %eax ## encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X64-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X64-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
+; X64-NEXT:    retq ## encoding: [0xc3]
   %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
   %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 -1)
@@ -6375,28 +6467,44 @@
 define <8 x i8> @test_mask_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; X86-LABEL: test_mask_ucmp_q_512:
 ; X86:       ## %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
+; X86-NEXT:    pushl %ebp ## encoding: [0x55]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %ebx ## encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %edi ## encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    pushl %esi ## encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 20
+; X86-NEXT:    .cfi_offset %esi, -20
+; X86-NEXT:    .cfi_offset %edi, -16
+; X86-NEXT:    .cfi_offset %ebx, -12
+; X86-NEXT:    .cfi_offset %ebp, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x14]
 ; X86-NEXT:    kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x29,0xc1]
 ; X86-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
 ; X86-NEXT:    vpcmpltuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x01]
-; X86-NEXT:    vpcmpleuq %zmm1, %zmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xd1,0x02]
-; X86-NEXT:    vpcmpneqq %zmm1, %zmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xd9,0x04]
-; X86-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xe1,0x05]
-; X86-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc9,0x06]
 ; X86-NEXT:    kmovw %k0, %edx ## encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpleuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %edi ## encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %ebx ## encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x06]
+; X86-NEXT:    kmovw %k0, %ebp ## encoding: [0xc5,0xf8,0x93,0xe8]
 ; X86-NEXT:    movzbl %cl, %ecx ## encoding: [0x0f,0xb6,0xc9]
 ; X86-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; X86-NEXT:    vpinsrb $1, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x01]
-; X86-NEXT:    kmovw %k2, %ecx ## encoding: [0xc5,0xf8,0x93,0xca]
-; X86-NEXT:    vpinsrb $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x02]
-; X86-NEXT:    kmovw %k3, %ecx ## encoding: [0xc5,0xf8,0x93,0xcb]
-; X86-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x04]
-; X86-NEXT:    kmovw %k4, %ecx ## encoding: [0xc5,0xf8,0x93,0xcc]
-; X86-NEXT:    vpinsrb $5, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x05]
-; X86-NEXT:    kmovw %k1, %ecx ## encoding: [0xc5,0xf8,0x93,0xc9]
-; X86-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
+; X86-NEXT:    vpinsrb $2, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x02]
+; X86-NEXT:    vpinsrb $4, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x04]
+; X86-NEXT:    vpinsrb $5, %ebx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x05]
+; X86-NEXT:    vpinsrb $6, %ebp, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc5,0x06]
 ; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi ## encoding: [0x5e]
+; X86-NEXT:    popl %edi ## encoding: [0x5f]
+; X86-NEXT:    popl %ebx ## encoding: [0x5b]
+; X86-NEXT:    popl %ebp ## encoding: [0x5d]
 ; X86-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
 ; X86-NEXT:    retl ## encoding: [0xc3]
 ;
@@ -6404,24 +6512,24 @@
 ; X64:       ## %bb.0:
 ; X64-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x29,0xc1]
-; X64-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    kmovw %k0, %r8d ## encoding: [0xc5,0x78,0x93,0xc0]
 ; X64-NEXT:    vpcmpltuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x01]
-; X64-NEXT:    vpcmpleuq %zmm1, %zmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xd1,0x02]
-; X64-NEXT:    vpcmpneqq %zmm1, %zmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xd9,0x04]
-; X64-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xe1,0x05]
-; X64-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc9,0x06]
+; X64-NEXT:    kmovw %k0, %r9d ## encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpleuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %r10d ## encoding: [0xc5,0x78,0x93,0xd0]
+; X64-NEXT:    vpcmpneqq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi ## encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x1e,0xc1,0x06]
 ; X64-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; X64-NEXT:    movzbl %al, %eax ## encoding: [0x0f,0xb6,0xc0]
-; X64-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; X64-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; X64-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; X64-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; X64-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; X64-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; X64-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; X64-NEXT:    movzbl %r8b, %edx ## encoding: [0x41,0x0f,0xb6,0xd0]
+; X64-NEXT:    vmovd %edx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc2]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 ## encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %r10d, %xmm0, %xmm0 ## encoding: [0xc4,0xc3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
 ; X64-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; X64-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
 ; X64-NEXT:    vpinsrb $7, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x07]
 ; X64-NEXT:    vzeroupper ## encoding: [0xc5,0xf8,0x77]
 ; X64-NEXT:    retq ## encoding: [0xc3]
diff --git a/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
--- a/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
@@ -9492,31 +9492,72 @@
 declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)
 
 define <8 x i8> @test_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
-; CHECK-LABEL: test_cmp_d_256:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
-; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
-; CHECK-NEXT:    vpcmpgtd %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x66,0xc0]
-; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k1 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc9,0x02]
-; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd1,0x04]
-; CHECK-NEXT:    vpcmpnltd %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd9,0x05]
-; CHECK-NEXT:    vpcmpgtd %ymm1, %ymm0, %k4 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xe1]
-; CHECK-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; CHECK-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
-; 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 %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; 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:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
-; CHECK-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
-; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
-; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
+; X86-LABEL: test_cmp_d_256:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebx # encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %edi # encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %esi # encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    .cfi_offset %esi, -16
+; X86-NEXT:    .cfi_offset %edi, -12
+; X86-NEXT:    .cfi_offset %ebx, -8
+; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
+; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X86-NEXT:    vpcmpgtd %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x66,0xc0]
+; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
+; X86-NEXT:    vpcmpled %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpnltd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %edi # encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xc1]
+; X86-NEXT:    kmovw %k0, %ebx # encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
+; X86-NEXT:    vmovd %eax, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; X86-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
+; X86-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X86-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X86-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X86-NEXT:    vpinsrb $6, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x06]
+; X86-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi # encoding: [0x5e]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    popl %edi # encoding: [0x5f]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    popl %ebx # encoding: [0x5b]
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_cmp_d_256:
+; X64:       # %bb.0:
+; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
+; X64-NEXT:    kmovw %k0, %r8d # encoding: [0xc5,0x78,0x93,0xc0]
+; X64-NEXT:    vpcmpgtd %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x66,0xc0]
+; X64-NEXT:    kmovw %k0, %r9d # encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpled %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
+; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %edi # encoding: [0xc5,0xf8,0x93,0xf8]
+; X64-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xc1]
+; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    movzbl %r8b, %ecx # encoding: [0x41,0x0f,0xb6,0xc8]
+; X64-NEXT:    vmovd %ecx, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X64-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X64-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X64-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)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
   %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
@@ -9539,28 +9580,48 @@
 define <8 x i8> @test_mask_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
 ; X86-LABEL: test_mask_cmp_d_256:
 ; X86:       # %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
+; X86-NEXT:    pushl %ebp # encoding: [0x55]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %ebx # encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %edi # encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    pushl %esi # encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 20
+; X86-NEXT:    .cfi_offset %esi, -20
+; X86-NEXT:    .cfi_offset %edi, -16
+; X86-NEXT:    .cfi_offset %ebx, -12
+; X86-NEXT:    .cfi_offset %ebp, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
 ; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
 ; X86-NEXT:    vpcmpgtd %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x66,0xc0]
-; X86-NEXT:    vpcmpled %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd1,0x02]
-; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x04]
-; X86-NEXT:    vpcmpnltd %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x05]
-; X86-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc9]
 ; X86-NEXT:    kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpled %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %edi # encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpnltd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %ebx # encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc1]
+; X86-NEXT:    kmovw %k0, %ebp # encoding: [0xc5,0xf8,0x93,0xe8]
 ; X86-NEXT:    movzbl %cl, %ecx # encoding: [0x0f,0xb6,0xc9]
 ; X86-NEXT:    vmovd %ecx, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; X86-NEXT:    vpinsrb $1, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x01]
-; X86-NEXT:    kmovw %k2, %ecx # encoding: [0xc5,0xf8,0x93,0xca]
-; X86-NEXT:    vpinsrb $2, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x02]
-; X86-NEXT:    kmovw %k3, %ecx # encoding: [0xc5,0xf8,0x93,0xcb]
-; X86-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x04]
-; X86-NEXT:    kmovw %k4, %ecx # encoding: [0xc5,0xf8,0x93,0xcc]
-; X86-NEXT:    vpinsrb $5, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x05]
-; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
-; X86-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
+; X86-NEXT:    vpinsrb $2, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x02]
+; X86-NEXT:    vpinsrb $4, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x04]
+; X86-NEXT:    vpinsrb $5, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x05]
+; X86-NEXT:    vpinsrb $6, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc5,0x06]
 ; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi # encoding: [0x5e]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    popl %edi # encoding: [0x5f]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    popl %ebx # encoding: [0x5b]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    popl %ebp # encoding: [0x5d]
+; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
@@ -9568,24 +9629,24 @@
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
-; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    kmovw %k0, %r8d # encoding: [0xc5,0x78,0x93,0xc0]
 ; X64-NEXT:    vpcmpgtd %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x66,0xc0]
-; X64-NEXT:    vpcmpled %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd1,0x02]
-; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x04]
-; X64-NEXT:    vpcmpnltd %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x05]
-; X64-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc9]
+; X64-NEXT:    kmovw %k0, %r9d # encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpled %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %r10d # encoding: [0xc5,0x78,0x93,0xd0]
+; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc1]
 ; X64-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
-; X64-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
-; X64-NEXT:    vmovd %eax, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; X64-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
-; X64-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
-; X64-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
+; X64-NEXT:    movzbl %r8b, %edx # encoding: [0x41,0x0f,0xb6,0xd0]
+; X64-NEXT:    vmovd %edx, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc2]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %r10d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
 ; X64-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
-; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
 ; X64-NEXT:    vpinsrb $7, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x07]
 ; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X64-NEXT:    retq # encoding: [0xc3]
@@ -9611,31 +9672,72 @@
 declare i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
 
 define <8 x i8> @test_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
-; CHECK-LABEL: test_ucmp_d_256:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
-; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
-; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x01]
-; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k1 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc9,0x02]
-; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd1,0x04]
-; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xd9,0x05]
-; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe1,0x06]
-; CHECK-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
-; CHECK-NEXT:    vmovd %eax, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; CHECK-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
-; 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 %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; 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:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
-; CHECK-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
-; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
-; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
+; X86-LABEL: test_ucmp_d_256:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %ebx # encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %edi # encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %esi # encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    .cfi_offset %esi, -16
+; X86-NEXT:    .cfi_offset %edi, -12
+; X86-NEXT:    .cfi_offset %ebx, -8
+; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
+; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X86-NEXT:    vpcmpltud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x01]
+; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
+; X86-NEXT:    vpcmpleud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpnltud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %edi # encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpnleud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x06]
+; X86-NEXT:    kmovw %k0, %ebx # encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
+; X86-NEXT:    vmovd %eax, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; X86-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
+; X86-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X86-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X86-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X86-NEXT:    vpinsrb $6, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x06]
+; X86-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi # encoding: [0x5e]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    popl %edi # encoding: [0x5f]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    popl %ebx # encoding: [0x5b]
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_ucmp_d_256:
+; X64:       # %bb.0:
+; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
+; X64-NEXT:    kmovw %k0, %r8d # encoding: [0xc5,0x78,0x93,0xc0]
+; X64-NEXT:    vpcmpltud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x01]
+; X64-NEXT:    kmovw %k0, %r9d # encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpleud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
+; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %edi # encoding: [0xc5,0xf8,0x93,0xf8]
+; X64-NEXT:    vpcmpnleud %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x06]
+; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    movzbl %r8b, %ecx # encoding: [0x41,0x0f,0xb6,0xc8]
+; X64-NEXT:    vmovd %ecx, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
+; X64-NEXT:    vpinsrb $5, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x05]
+; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
+; X64-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X64-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)
   %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
   %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
@@ -9658,28 +9760,48 @@
 define <8 x i8> @test_mask_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
 ; X86-LABEL: test_mask_ucmp_d_256:
 ; X86:       # %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
+; X86-NEXT:    pushl %ebp # encoding: [0x55]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    pushl %ebx # encoding: [0x53]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    pushl %edi # encoding: [0x57]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    pushl %esi # encoding: [0x56]
+; X86-NEXT:    .cfi_def_cfa_offset 20
+; X86-NEXT:    .cfi_offset %esi, -20
+; X86-NEXT:    .cfi_offset %edi, -16
+; X86-NEXT:    .cfi_offset %ebx, -12
+; X86-NEXT:    .cfi_offset %ebp, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
 ; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
 ; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
 ; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
 ; X86-NEXT:    vpcmpltud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x01]
-; X86-NEXT:    vpcmpleud %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x02]
-; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x04]
-; X86-NEXT:    vpcmpnltud %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe1,0x05]
-; X86-NEXT:    vpcmpnleud %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x06]
 ; X86-NEXT:    kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
+; X86-NEXT:    vpcmpleud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x02]
+; X86-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x04]
+; X86-NEXT:    kmovw %k0, %edi # encoding: [0xc5,0xf8,0x93,0xf8]
+; X86-NEXT:    vpcmpnltud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x05]
+; X86-NEXT:    kmovw %k0, %ebx # encoding: [0xc5,0xf8,0x93,0xd8]
+; X86-NEXT:    vpcmpnleud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x06]
+; X86-NEXT:    kmovw %k0, %ebp # encoding: [0xc5,0xf8,0x93,0xe8]
 ; X86-NEXT:    movzbl %cl, %ecx # encoding: [0x0f,0xb6,0xc9]
 ; X86-NEXT:    vmovd %ecx, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
 ; X86-NEXT:    vpinsrb $1, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x01]
-; X86-NEXT:    kmovw %k2, %ecx # encoding: [0xc5,0xf8,0x93,0xca]
-; X86-NEXT:    vpinsrb $2, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x02]
-; X86-NEXT:    kmovw %k3, %ecx # encoding: [0xc5,0xf8,0x93,0xcb]
-; X86-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x04]
-; X86-NEXT:    kmovw %k4, %ecx # encoding: [0xc5,0xf8,0x93,0xcc]
-; X86-NEXT:    vpinsrb $5, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x05]
-; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
-; X86-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
+; X86-NEXT:    vpinsrb $2, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x02]
+; X86-NEXT:    vpinsrb $4, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x04]
+; X86-NEXT:    vpinsrb $5, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc3,0x05]
+; X86-NEXT:    vpinsrb $6, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc5,0x06]
 ; X86-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
+; X86-NEXT:    popl %esi # encoding: [0x5e]
+; X86-NEXT:    .cfi_def_cfa_offset 16
+; X86-NEXT:    popl %edi # encoding: [0x5f]
+; X86-NEXT:    .cfi_def_cfa_offset 12
+; X86-NEXT:    popl %ebx # encoding: [0x5b]
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    popl %ebp # encoding: [0x5d]
+; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
@@ -9687,24 +9809,24 @@
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
-; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    kmovw %k0, %r8d # encoding: [0xc5,0x78,0x93,0xc0]
 ; X64-NEXT:    vpcmpltud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x01]
-; X64-NEXT:    vpcmpleud %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x02]
-; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x04]
-; X64-NEXT:    vpcmpnltud %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe1,0x05]
-; X64-NEXT:    vpcmpnleud %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x06]
+; X64-NEXT:    kmovw %k0, %r9d # encoding: [0xc5,0x78,0x93,0xc8]
+; X64-NEXT:    vpcmpleud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x02]
+; X64-NEXT:    kmovw %k0, %r10d # encoding: [0xc5,0x78,0x93,0xd0]
+; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x04]
+; X64-NEXT:    kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
+; X64-NEXT:    vpcmpnltud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x05]
+; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
+; X64-NEXT:    vpcmpnleud %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x06]
 ; X64-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
-; X64-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
-; X64-NEXT:    vmovd %eax, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
-; X64-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
-; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
-; X64-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
-; X64-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
+; X64-NEXT:    movzbl %r8b, %edx # encoding: [0x41,0x0f,0xb6,0xd0]
+; X64-NEXT:    vmovd %edx, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc2]
+; X64-NEXT:    vpinsrb $1, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x01]
+; X64-NEXT:    vpinsrb $2, %r10d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc2,0x02]
+; X64-NEXT:    vpinsrb $4, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x04]
 ; X64-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
-; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
-; X64-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; X64-NEXT:    vpinsrb $6, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
 ; X64-NEXT:    vpinsrb $7, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x07]
 ; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X64-NEXT:    retq # encoding: [0xc3]
diff --git a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
--- a/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
+++ b/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool.ll
@@ -147,8 +147,7 @@
 ;
 ; AVX1-LABEL: bitcast_i8_8i1_freeze:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    movzbl %dil, %eax
-; AVX1-NEXT:    vmovd %eax, %xmm0
+; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128]
diff --git a/llvm/test/CodeGen/X86/load-scalar-as-vector.ll b/llvm/test/CodeGen/X86/load-scalar-as-vector.ll
--- a/llvm/test/CodeGen/X86/load-scalar-as-vector.ll
+++ b/llvm/test/CodeGen/X86/load-scalar-as-vector.ll
@@ -95,14 +95,6 @@
 ; SSE-NEXT:    movzbl %al, %eax
 ; SSE-NEXT:    movd %eax, %xmm0
 ; SSE-NEXT:    retq
-;
-; AVX-LABEL: sub_op1_constant:
-; AVX:       # %bb.0:
-; AVX-NEXT:    movzbl (%rdi), %eax
-; AVX-NEXT:    addb $-42, %al
-; AVX-NEXT:    movzbl %al, %eax
-; AVX-NEXT:    vmovd %eax, %xmm0
-; AVX-NEXT:    retq
   %x = load i8, ptr %p
   %b = sub i8 %x, 42
   %r = insertelement <16 x i8> undef, i8 %b, i32 0
@@ -237,14 +229,6 @@
 ; SSE-NEXT:    movzbl %al, %eax
 ; SSE-NEXT:    movd %eax, %xmm0
 ; SSE-NEXT:    retq
-;
-; AVX-LABEL: shl_op1_constant:
-; AVX:       # %bb.0:
-; AVX-NEXT:    movzbl (%rdi), %eax
-; AVX-NEXT:    shlb $5, %al
-; AVX-NEXT:    movzbl %al, %eax
-; AVX-NEXT:    vmovd %eax, %xmm0
-; AVX-NEXT:    retq
   %x = load i8, ptr %p
   %b = shl i8 %x, 5
   %r = insertelement <16 x i8> undef, i8 %b, i32 0
@@ -531,20 +515,6 @@
 ; SSE-NEXT:    movzbl %al, %eax
 ; SSE-NEXT:    movd %eax, %xmm0
 ; SSE-NEXT:    retq
-;
-; AVX-LABEL: urem_op1_constant:
-; AVX:       # %bb.0:
-; AVX-NEXT:    movzbl (%rdi), %eax
-; AVX-NEXT:    movl %eax, %ecx
-; AVX-NEXT:    shrb %cl
-; AVX-NEXT:    movzbl %cl, %ecx
-; AVX-NEXT:    imull $49, %ecx, %ecx
-; AVX-NEXT:    shrl $10, %ecx
-; AVX-NEXT:    imull $42, %ecx, %ecx
-; AVX-NEXT:    subb %cl, %al
-; AVX-NEXT:    movzbl %al, %eax
-; AVX-NEXT:    vmovd %eax, %xmm0
-; AVX-NEXT:    retq
   %x = load i8, ptr %p
   %b = urem i8 %x, 42
   %r = insertelement <16 x i8> undef, i8 %b, i32 0
diff --git a/llvm/test/CodeGen/X86/pr15267.ll b/llvm/test/CodeGen/X86/pr15267.ll
--- a/llvm/test/CodeGen/X86/pr15267.ll
+++ b/llvm/test/CodeGen/X86/pr15267.ll
@@ -31,19 +31,15 @@
 ; CHECK-NEXT:    movl %eax, %ecx
 ; CHECK-NEXT:    shrb %cl
 ; CHECK-NEXT:    andb $1, %cl
-; CHECK-NEXT:    movzbl %cl, %ecx
 ; CHECK-NEXT:    movl %eax, %edx
 ; CHECK-NEXT:    andb $1, %dl
-; CHECK-NEXT:    movzbl %dl, %edx
 ; CHECK-NEXT:    vmovd %edx, %xmm0
 ; CHECK-NEXT:    vpinsrb $4, %ecx, %xmm0, %xmm0
 ; CHECK-NEXT:    movl %eax, %ecx
 ; CHECK-NEXT:    shrb $2, %cl
 ; CHECK-NEXT:    andb $1, %cl
-; CHECK-NEXT:    movzbl %cl, %ecx
 ; CHECK-NEXT:    vpinsrb $8, %ecx, %xmm0, %xmm0
 ; CHECK-NEXT:    shrb $3, %al
-; CHECK-NEXT:    movzbl %al, %eax
 ; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %ret = load <4 x i1>, ptr %in, align 1
diff --git a/llvm/test/CodeGen/X86/setcc-lowering.ll b/llvm/test/CodeGen/X86/setcc-lowering.ll
--- a/llvm/test/CodeGen/X86/setcc-lowering.ll
+++ b/llvm/test/CodeGen/X86/setcc-lowering.ll
@@ -44,7 +44,6 @@
 ; AVX-NEXT:    .p2align 4, 0x90
 ; AVX-NEXT:  .LBB1_1: # %for_loop599
 ; AVX-NEXT:    # =>This Inner Loop Header: Depth=1
-; AVX-NEXT:    xorl %eax, %eax
 ; AVX-NEXT:    cmpq $65536, %rdi # imm = 0x10000
 ; AVX-NEXT:    setl %al
 ; AVX-NEXT:    vmovd %eax, %xmm2
diff --git a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
@@ -3395,9 +3395,9 @@
 ;
 ; X86-AVX1-LABEL: test_mm_set_epi8:
 ; X86-AVX1:       # %bb.0:
+; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x40]
+; X86-AVX1-NEXT:    vmovd %eax, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc0]
 ; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x3c]
-; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x40]
-; X86-AVX1-NEXT:    vmovd %ecx, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc1]
 ; X86-AVX1-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
 ; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x38]
 ; X86-AVX1-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -3533,10 +3533,10 @@
 ;
 ; X64-AVX1-LABEL: test_mm_set_epi8:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %r10d # encoding: [0x44,0x0f,0xb6,0x54,0x24,0x48]
 ; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x50]
 ; X64-AVX1-NEXT:    vmovd %eax, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc0]
-; X64-AVX1-NEXT:    vpinsrb $1, %r10d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc2,0x01]
+; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x48]
+; X64-AVX1-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
 ; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x40]
 ; X64-AVX1-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
 ; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x38]
@@ -3553,18 +3553,12 @@
 ; X64-AVX1-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
 ; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X64-AVX1-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x09]
-; X64-AVX1-NEXT:    movzbl %r9b, %eax # encoding: [0x41,0x0f,0xb6,0xc1]
-; X64-AVX1-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; X64-AVX1-NEXT:    movzbl %r8b, %eax # encoding: [0x41,0x0f,0xb6,0xc0]
-; X64-AVX1-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0b]
-; X64-AVX1-NEXT:    movzbl %cl, %eax # encoding: [0x0f,0xb6,0xc1]
-; X64-AVX1-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; X64-AVX1-NEXT:    movzbl %dl, %eax # encoding: [0x0f,0xb6,0xc2]
-; X64-AVX1-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0d]
-; X64-AVX1-NEXT:    movzbl %sil, %eax # encoding: [0x40,0x0f,0xb6,0xc6]
-; X64-AVX1-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; X64-AVX1-NEXT:    movzbl %dil, %eax # encoding: [0x40,0x0f,0xb6,0xc7]
-; X64-AVX1-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0f]
+; X64-AVX1-NEXT:    vpinsrb $10, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x0a]
+; X64-AVX1-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc0,0x0b]
+; X64-AVX1-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x0c]
+; X64-AVX1-NEXT:    vpinsrb $13, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x0d]
+; X64-AVX1-NEXT:    vpinsrb $14, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x0e]
+; X64-AVX1-NEXT:    vpinsrb $15, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0f]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
 ;
 ; X64-AVX512-LABEL: test_mm_set_epi8:
@@ -3671,10 +3665,10 @@
 ;
 ; X32-AVX1-LABEL: test_mm_set_epi8:
 ; X32-AVX1:       # %bb.0:
-; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %r10d # encoding: [0x67,0x44,0x0f,0xb6,0x54,0x24,0x48]
 ; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x50]
 ; X32-AVX1-NEXT:    vmovd %eax, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc0]
-; X32-AVX1-NEXT:    vpinsrb $1, %r10d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc2,0x01]
+; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x48]
+; X32-AVX1-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
 ; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x40]
 ; X32-AVX1-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
 ; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x38]
@@ -3691,18 +3685,12 @@
 ; X32-AVX1-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
 ; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x08]
 ; X32-AVX1-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x09]
-; X32-AVX1-NEXT:    movzbl %r9b, %eax # encoding: [0x41,0x0f,0xb6,0xc1]
-; X32-AVX1-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; X32-AVX1-NEXT:    movzbl %r8b, %eax # encoding: [0x41,0x0f,0xb6,0xc0]
-; X32-AVX1-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0b]
-; X32-AVX1-NEXT:    movzbl %cl, %eax # encoding: [0x0f,0xb6,0xc1]
-; X32-AVX1-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; X32-AVX1-NEXT:    movzbl %dl, %eax # encoding: [0x0f,0xb6,0xc2]
-; X32-AVX1-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0d]
-; X32-AVX1-NEXT:    movzbl %sil, %eax # encoding: [0x40,0x0f,0xb6,0xc6]
-; X32-AVX1-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; X32-AVX1-NEXT:    movzbl %dil, %eax # encoding: [0x40,0x0f,0xb6,0xc7]
-; X32-AVX1-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0f]
+; X32-AVX1-NEXT:    vpinsrb $10, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x0a]
+; X32-AVX1-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc0,0x0b]
+; X32-AVX1-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x0c]
+; X32-AVX1-NEXT:    vpinsrb $13, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x0d]
+; X32-AVX1-NEXT:    vpinsrb $14, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x0e]
+; X32-AVX1-NEXT:    vpinsrb $15, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0f]
 ; X32-AVX1-NEXT:    retq # encoding: [0xc3]
 ;
 ; X32-AVX512-LABEL: test_mm_set_epi8:
@@ -4397,8 +4385,7 @@
 ;
 ; X64-AVX1-LABEL: test_mm_set1_epi8:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    movzbl %dil, %eax # encoding: [0x40,0x0f,0xb6,0xc7]
-; X64-AVX1-NEXT:    vmovd %eax, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc0]
+; X64-AVX1-NEXT:    vmovd %edi, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc7]
 ; X64-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0xef,0xc9]
 ; X64-AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe2,0x79,0x00,0xc1]
 ; X64-AVX1-NEXT:    retq # encoding: [0xc3]
@@ -4422,8 +4409,7 @@
 ;
 ; X32-AVX1-LABEL: test_mm_set1_epi8:
 ; X32-AVX1:       # %bb.0:
-; X32-AVX1-NEXT:    movzbl %dil, %eax # encoding: [0x40,0x0f,0xb6,0xc7]
-; X32-AVX1-NEXT:    vmovd %eax, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc0]
+; X32-AVX1-NEXT:    vmovd %edi, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc7]
 ; X32-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0xef,0xc9]
 ; X32-AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe2,0x79,0x00,0xc1]
 ; X32-AVX1-NEXT:    retq # encoding: [0xc3]
@@ -4812,9 +4798,9 @@
 ;
 ; X86-AVX1-LABEL: test_mm_setr_epi8:
 ; X86-AVX1:       # %bb.0:
+; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
+; X86-AVX1-NEXT:    vmovd %eax, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc0]
 ; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
-; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x04]
-; X86-AVX1-NEXT:    vmovd %ecx, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc1]
 ; X86-AVX1-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
 ; X86-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x0c]
 ; X86-AVX1-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -4950,18 +4936,12 @@
 ;
 ; X64-AVX1-LABEL: test_mm_setr_epi8:
 ; X64-AVX1:       # %bb.0:
-; X64-AVX1-NEXT:    movzbl %sil, %eax # encoding: [0x40,0x0f,0xb6,0xc6]
-; X64-AVX1-NEXT:    movzbl %dil, %esi # encoding: [0x40,0x0f,0xb6,0xf7]
-; X64-AVX1-NEXT:    vmovd %esi, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc6]
-; X64-AVX1-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
-; X64-AVX1-NEXT:    movzbl %dl, %eax # encoding: [0x0f,0xb6,0xc2]
-; X64-AVX1-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; X64-AVX1-NEXT:    movzbl %cl, %eax # encoding: [0x0f,0xb6,0xc1]
-; X64-AVX1-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x03]
-; X64-AVX1-NEXT:    movzbl %r8b, %eax # encoding: [0x41,0x0f,0xb6,0xc0]
-; X64-AVX1-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; X64-AVX1-NEXT:    movzbl %r9b, %eax # encoding: [0x41,0x0f,0xb6,0xc1]
-; X64-AVX1-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
+; X64-AVX1-NEXT:    vmovd %edi, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc7]
+; X64-AVX1-NEXT:    vpinsrb $1, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x01]
+; X64-AVX1-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X64-AVX1-NEXT:    vpinsrb $3, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x03]
+; X64-AVX1-NEXT:    vpinsrb $4, %r8d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc0,0x04]
+; X64-AVX1-NEXT:    vpinsrb $5, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x05]
 ; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
 ; X64-AVX1-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
 ; X64-AVX1-NEXT:    movzbl {{[0-9]+}}(%rsp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x10]
@@ -5088,18 +5068,12 @@
 ;
 ; X32-AVX1-LABEL: test_mm_setr_epi8:
 ; X32-AVX1:       # %bb.0:
-; X32-AVX1-NEXT:    movzbl %sil, %eax # encoding: [0x40,0x0f,0xb6,0xc6]
-; X32-AVX1-NEXT:    movzbl %dil, %esi # encoding: [0x40,0x0f,0xb6,0xf7]
-; X32-AVX1-NEXT:    vmovd %esi, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc6]
-; X32-AVX1-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
-; X32-AVX1-NEXT:    movzbl %dl, %eax # encoding: [0x0f,0xb6,0xc2]
-; X32-AVX1-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; X32-AVX1-NEXT:    movzbl %cl, %eax # encoding: [0x0f,0xb6,0xc1]
-; X32-AVX1-NEXT:    vpinsrb $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x03]
-; X32-AVX1-NEXT:    movzbl %r8b, %eax # encoding: [0x41,0x0f,0xb6,0xc0]
-; X32-AVX1-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; X32-AVX1-NEXT:    movzbl %r9b, %eax # encoding: [0x41,0x0f,0xb6,0xc1]
-; X32-AVX1-NEXT:    vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
+; X32-AVX1-NEXT:    vmovd %edi, %xmm0 # encoding: [0xc5,0xf9,0x6e,0xc7]
+; X32-AVX1-NEXT:    vpinsrb $1, %esi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc6,0x01]
+; X32-AVX1-NEXT:    vpinsrb $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc2,0x02]
+; X32-AVX1-NEXT:    vpinsrb $3, %ecx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x03]
+; X32-AVX1-NEXT:    vpinsrb $4, %r8d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc0,0x04]
+; X32-AVX1-NEXT:    vpinsrb $5, %r9d, %xmm0, %xmm0 # encoding: [0xc4,0xc3,0x79,0x20,0xc1,0x05]
 ; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x08]
 ; X32-AVX1-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
 ; X32-AVX1-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x67,0x0f,0xb6,0x44,0x24,0x10]
diff --git a/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll
--- a/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll
@@ -571,12 +571,6 @@
 ; X64-SSE-NEXT:    movzbl %dil, %eax
 ; X64-SSE-NEXT:    pinsrb $1, %eax, %xmm0
 ; X64-SSE-NEXT:    retq
-;
-; X64-AVX-LABEL: test_mm_insert_epi8:
-; X64-AVX:       # %bb.0:
-; X64-AVX-NEXT:    movzbl %dil, %eax
-; X64-AVX-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
-; X64-AVX-NEXT:    retq
   %arg0 = bitcast <2 x i64> %a0 to <16 x i8>
   %res = insertelement <16 x i8> %arg0, i8 %a1,i32 1
   %bc = bitcast <16 x i8> %res to <2 x i64>