Index: llvm/trunk/lib/Target/X86/X86InstrAVX512.td
===================================================================
--- llvm/trunk/lib/Target/X86/X86InstrAVX512.td
+++ llvm/trunk/lib/Target/X86/X86InstrAVX512.td
@@ -1242,8 +1242,9 @@
              (ins _.KRCWM:$mask, _.RC:$src1, _.RC:$src2),
              !strconcat(OpcodeStr,
              "\t{$src2, $src1, ${dst} {${mask}}|${dst} {${mask}}, $src1, $src2}"),
-             [(set _.RC:$dst, (X86select _.KRCWM:$mask, (_.VT _.RC:$src1),
-                 (_.VT _.RC:$src2)))]>, EVEX_4V, EVEX_K;
+             [(set _.RC:$dst, (vselect _.KRCWM:$mask, 
+                                (_.VT _.RC:$src2),
+                                (_.VT _.RC:$src1)))]>, EVEX_4V, EVEX_K;
   let hasSideEffects = 0 in
   def rrkz : AVX5128I<opc, MRMSrcReg, (outs _.RC:$dst),
              (ins _.KRCWM:$mask, _.RC:$src1, _.RC:$src2),
@@ -1260,8 +1261,9 @@
              (ins _.KRCWM:$mask, _.RC:$src1, _.MemOp:$src2),
              !strconcat(OpcodeStr,
              "\t{$src2, $src1, ${dst} {${mask}}|${dst} {${mask}}, $src1, $src2}"),
-             [(set _.RC:$dst, (X86select _.KRCWM:$mask, (_.VT _.RC:$src1),
-              (_.VT (bitconvert (_.LdFrag addr:$src2)))))]>,
+             [(set _.RC:$dst, (vselect _.KRCWM:$mask,
+                                 (_.VT (bitconvert (_.LdFrag addr:$src2))),
+                                 (_.VT _.RC:$src1)))]>,
               EVEX_4V, EVEX_K, EVEX_CD8<_.EltSize, CD8VF>;
   let mayLoad = 1, hasSideEffects = 0 in
   def rmkz : AVX5128I<opc, MRMSrcMem, (outs _.RC:$dst),
@@ -1278,8 +1280,9 @@
        !strconcat(OpcodeStr,
             "\t{${src2}", _.BroadcastStr, ", $src1, $dst {${mask}}|",
             "$dst {${mask}}, $src1, ${src2}", _.BroadcastStr, "}"),
-      [(set _.RC:$dst,(X86select _.KRCWM:$mask, (_.VT _.RC:$src1),
-                       (X86VBroadcast (_.ScalarLdFrag addr:$src2))))]>,
+      [(set _.RC:$dst,(vselect _.KRCWM:$mask,
+                        (X86VBroadcast (_.ScalarLdFrag addr:$src2)),
+                        (_.VT _.RC:$src1)))]>,
       EVEX_4V, EVEX_K, EVEX_B, EVEX_CD8<_.EltSize, CD8VF>;
 
   let mayLoad = 1, hasSideEffects = 0 in
Index: llvm/trunk/test/CodeGen/X86/avx512-bugfix-23634.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-bugfix-23634.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-bugfix-23634.ll
@@ -15,7 +15,7 @@
 ; CHECK-NEXT:    vpsrad $1, %zmm2, %zmm2
 ; CHECK-NEXT:    movw $-21846, %ax ## imm = 0xAAAA
 ; CHECK-NEXT:    kmovw %eax, %k1
-; CHECK-NEXT:    vmovdqa32 {{.*}}(%rip), %zmm1 {%k1}
+; CHECK-NEXT:    vpblendmd {{.*}}(%rip), %zmm1, %zmm1 {%k1}
 ; CHECK-NEXT:    vpaddd %zmm0, %zmm2, %zmm0
 ; CHECK-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    vcvtdq2ps %zmm0, %zmm0
Index: llvm/trunk/test/CodeGen/X86/avx512-mov.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-mov.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-mov.ll
@@ -313,7 +313,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x48,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqa32 (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0x6f,0x07]
+; CHECK-NEXT:    vpblendmd (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x49,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <16 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <16 x i32>*
@@ -327,7 +327,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x48,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu32 (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7e,0x49,0x6f,0x07]
+; CHECK-NEXT:    vpblendmd (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x49,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <16 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <16 x i32>*
@@ -369,7 +369,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x48,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0x6f,0x07]
+; CHECK-NEXT:    vpblendmq (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <8 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x i64>*
@@ -383,7 +383,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x48,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu64 (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfe,0x49,0x6f,0x07]
+; CHECK-NEXT:    vpblendmq (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <8 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x i64>*
@@ -426,7 +426,7 @@
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vcmpordps %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0x74,0x48,0xc2,0xca,0x07]
 ; CHECK-NEXT:    vcmpneqps %zmm2, %zmm1, %k1 {%k1} ## encoding: [0x62,0xf1,0x74,0x49,0xc2,0xca,0x04]
-; CHECK-NEXT:    vmovaps (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0x28,0x07]
+; CHECK-NEXT:    vblendmps (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x49,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <16 x float>*
@@ -441,7 +441,7 @@
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vcmpordps %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0x74,0x48,0xc2,0xca,0x07]
 ; CHECK-NEXT:    vcmpneqps %zmm2, %zmm1, %k1 {%k1} ## encoding: [0x62,0xf1,0x74,0x49,0xc2,0xca,0x04]
-; CHECK-NEXT:    vmovups (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0x10,0x07]
+; CHECK-NEXT:    vblendmps (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x49,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <16 x float>*
@@ -486,7 +486,7 @@
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vcmpordpd %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0xf5,0x48,0xc2,0xca,0x07]
 ; CHECK-NEXT:    vcmpneqpd %zmm2, %zmm1, %k1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xc2,0xca,0x04]
-; CHECK-NEXT:    vmovapd (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0x28,0x07]
+; CHECK-NEXT:    vblendmpd (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x double>*
@@ -501,7 +501,7 @@
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2 ## encoding: [0x62,0xf1,0x6d,0x48,0xef,0xd2]
 ; CHECK-NEXT:    vcmpordpd %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0xf5,0x48,0xc2,0xca,0x07]
 ; CHECK-NEXT:    vcmpneqpd %zmm2, %zmm1, %k1 {%k1} ## encoding: [0x62,0xf1,0xf5,0x49,0xc2,0xca,0x04]
-; CHECK-NEXT:    vmovupd (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0x10,0x07]
+; CHECK-NEXT:    vblendmpd (%rdi), %zmm0, %zmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x double>*
Index: llvm/trunk/test/CodeGen/X86/avx512-vec-cmp.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-vec-cmp.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-vec-cmp.ll
@@ -6,8 +6,7 @@
 ; CHECK-LABEL: test1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpleps %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = fcmp ole <16 x float> %x, %y
   %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %y
@@ -18,8 +17,7 @@
 ; CHECK-LABEL: test2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmplepd %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = fcmp ole <8 x double> %x, %y
   %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %y
@@ -30,8 +28,7 @@
 ; CHECK-LABEL: test3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i32>, <16 x i32>* %yp, align 4
   %mask = icmp eq <16 x i32> %x, %y
@@ -43,8 +40,7 @@
 ; CHECK-LABEL: test4_unsigned:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp uge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y
@@ -55,8 +51,7 @@
 ; CHECK-LABEL: test5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp eq <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x, <8 x i64> %y
@@ -67,8 +62,7 @@
 ; CHECK-LABEL: test6_unsigned:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ugt <8 x i64> %x, %y
   %max = select <8 x i1> %mask, <8 x i64> %x1, <8 x i64> %y
@@ -87,8 +81,7 @@
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxord %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vcmpltps %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovaps %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmps %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %mask = fcmp olt <4 x float> %a, zeroinitializer
@@ -108,8 +101,7 @@
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxord %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vcmpltpd %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovapd %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = fcmp olt <2 x double> %a, zeroinitializer
   %c = select <2 x i1>%mask, <2 x double>%a, <2 x double>%b
@@ -126,8 +118,7 @@
 ; SKX-LABEL: test9:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpcmpeqd %ymm1, %ymm0, %k1
-; SKX-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <8 x i32> %x, %y
   %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y
@@ -144,8 +135,7 @@
 ; SKX-LABEL: test10:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpeqps %ymm1, %ymm0, %k1
-; SKX-NEXT:    vmovaps %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %mask = fcmp oeq <8 x float> %x, %y
@@ -248,8 +238,7 @@
 ; CHECK-LABEL: test16:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sge <16 x i32> %x, %y
   %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y
@@ -260,8 +249,7 @@
 ; CHECK-LABEL: test17:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask = icmp sgt <16 x i32> %x, %y
@@ -273,8 +261,7 @@
 ; CHECK-LABEL: test18:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask = icmp sle <16 x i32> %x, %y
@@ -286,8 +273,7 @@
 ; CHECK-LABEL: test19:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
   %mask = icmp ule <16 x i32> %x, %y
@@ -300,8 +286,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpcmpeqd %zmm3, %zmm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp eq <16 x i32> %x1, %y1
   %mask0 = icmp eq <16 x i32> %x, %y
@@ -315,8 +300,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpcmpleq %zmm2, %zmm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmq %zmm0, %zmm2, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %mask0 = icmp sle <8 x i64> %x, %y
@@ -330,8 +314,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sgt <8 x i64> %x1, %y1
   %y = load <8 x i64>, <8 x i64>* %y.ptr, align 4
@@ -346,8 +329,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %zmm1, %zmm2, %k1
 ; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4
@@ -361,8 +343,7 @@
 ; CHECK-LABEL: test24:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to8}, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %yb = load i64, i64* %yb.ptr, align 4
   %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0
@@ -376,8 +357,7 @@
 ; CHECK-LABEL: test25:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi){1to16}, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %yb = load i32, i32* %yb.ptr, align 4
   %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0
@@ -392,8 +372,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %zmm1, %zmm2, %k1
 ; CHECK-NEXT:    vpcmpgtd (%rdi){1to16}, %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <16 x i32> %x1, %y1
   %yb = load i32, i32* %yb.ptr, align 4
@@ -410,8 +389,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %zmm1, %zmm2, %k1
 ; CHECK-NEXT:    vpcmpleq (%rdi){1to8}, %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i64> %x1, %y1
   %yb = load i64, i64* %yb.ptr, align 4
@@ -481,8 +459,7 @@
 ; SKX-LABEL: test30:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpeqpd %ymm1, %ymm0, %k1
-; SKX-NEXT:    vmovapd %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %mask = fcmp oeq <4 x double> %x, %y
@@ -500,8 +477,7 @@
 ; SKX-LABEL: test31:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi), %xmm0, %k1
-; SKX-NEXT:    vmovapd %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %y = load <2 x double>, <2 x double>* %yp, align 4
@@ -520,8 +496,7 @@
 ; SKX-LABEL: test32:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi), %ymm0, %k1
-; SKX-NEXT:    vmovapd %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %y = load <4 x double>, <4 x double>* %yp, align 4
@@ -534,8 +509,7 @@
 ; CHECK-LABEL: test33:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpltpd (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x double>, <8 x double>* %yp, align 4
   %mask = fcmp olt <8 x double> %x, %y
@@ -553,8 +527,7 @@
 ; SKX-LABEL: test34:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi), %xmm0, %k1
-; SKX-NEXT:    vmovaps %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmps %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
   %y = load <4 x float>, <4 x float>* %yp, align 4
   %mask = fcmp olt <4 x float> %x, %y
@@ -573,8 +546,7 @@
 ; SKX-LABEL: test35:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi), %ymm0, %k1
-; SKX-NEXT:    vmovaps %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %y = load <8 x float>, <8 x float>* %yp, align 4
@@ -587,8 +559,7 @@
 ; CHECK-LABEL: test36:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpltps (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x float>, <16 x float>* %yp, align 4
   %mask = fcmp olt <16 x float> %x, %y
@@ -600,8 +571,7 @@
 ; CHECK-LABEL: test37:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1
-; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
 
   %a = load double, double* %ptr
@@ -624,8 +594,7 @@
 ; SKX-LABEL: test38:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi){1to4}, %ymm0, %k1
-; SKX-NEXT:    vmovapd %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load double, double* %ptr
@@ -648,8 +617,7 @@
 ; SKX-LABEL: test39:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi){1to2}, %xmm0, %k1
-; SKX-NEXT:    vmovapd %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load double, double* %ptr
@@ -666,8 +634,7 @@
 ; CHECK-LABEL: test40:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpltps (%rdi){1to16}, %zmm0, %k1
-; CHECK-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
 
   %a = load float, float* %ptr
@@ -690,8 +657,7 @@
 ; SKX-LABEL: test41:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi){1to8}, %ymm0, %k1
-; SKX-NEXT:    vmovaps %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load float, float* %ptr
@@ -714,8 +680,7 @@
 ; SKX-LABEL: test42:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi){1to4}, %xmm0, %k1
-; SKX-NEXT:    vmovaps %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmps %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load float, float* %ptr
@@ -734,8 +699,7 @@
 ; KNL-NEXT:    vpsllq $63, %zmm2, %zmm2
 ; KNL-NEXT:    vptestmq %zmm2, %zmm2, %k1
 ; KNL-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1 {%k1}
-; KNL-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vmovaps %zmm1, %zmm0
+; KNL-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test43:
@@ -743,8 +707,7 @@
 ; SKX-NEXT:    vpsllw $15, %xmm2, %xmm2
 ; SKX-NEXT:    vpmovw2m %xmm2, %k1
 ; SKX-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1 {%k1}
-; SKX-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; SKX-NEXT:    vmovaps %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load double, double* %ptr
Index: llvm/trunk/test/CodeGen/X86/avx512bw-mov.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512bw-mov.ll
+++ llvm/trunk/test/CodeGen/X86/avx512bw-mov.ll
@@ -26,7 +26,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2
 ; CHECK-NEXT:    vpcmpneqb %zmm2, %zmm1, %k1
-; CHECK-NEXT:    vmovdqu8 (%rdi), %zmm0 {%k1}
+; CHECK-NEXT:    vpblendmb (%rdi), %zmm0, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ne <64 x i8> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <64 x i8>*
@@ -74,7 +74,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %zmm2, %zmm2, %zmm2
 ; CHECK-NEXT:    vpcmpneqw %zmm2, %zmm1, %k1
-; CHECK-NEXT:    vmovdqu16 (%rdi), %zmm0 {%k1}
+; CHECK-NEXT:    vpblendmw (%rdi), %zmm0, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ne <32 x i16> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <32 x i16>*
Index: llvm/trunk/test/CodeGen/X86/avx512bw-vec-cmp.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512bw-vec-cmp.ll
+++ llvm/trunk/test/CodeGen/X86/avx512bw-vec-cmp.ll
@@ -5,8 +5,7 @@
 ; CHECK-LABEL: test1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqb %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqu8 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp eq <64 x i8> %x, %y
   %max = select <64 x i1> %mask, <64 x i8> %x, <64 x i8> %y
@@ -17,8 +16,7 @@
 ; CHECK-LABEL: test2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtb %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqu8 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sgt <64 x i8> %x, %y
   %max = select <64 x i1> %mask, <64 x i8> %x1, <64 x i8> %y
@@ -29,8 +27,7 @@
 ; CHECK-LABEL: test3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew %zmm0, %zmm1, %k1
-; CHECK-NEXT:    vmovdqu16 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sge <32 x i16> %x, %y
   %max = select <32 x i1> %mask, <32 x i16> %x1, <32 x i16> %y
@@ -41,8 +38,7 @@
 ; CHECK-LABEL: test4:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleub %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vmovdqu8 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %zmm2, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ugt <64 x i8> %x, %y
   %max = select <64 x i1> %mask, <64 x i8> %x1, <64 x i8> %y
@@ -53,8 +49,7 @@
 ; CHECK-LABEL: test5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqw (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <32 x i16>, <32 x i16>* %yp, align 4
   %mask = icmp eq <32 x i16> %x, %y
@@ -66,8 +61,7 @@
 ; CHECK-LABEL: test6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtw (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <32 x i16>, <32 x i16>* %y.ptr, align 4
   %mask = icmp sgt <32 x i16> %x, %y
@@ -79,8 +73,7 @@
 ; CHECK-LABEL: test7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <32 x i16>, <32 x i16>* %y.ptr, align 4
   %mask = icmp sle <32 x i16> %x, %y
@@ -92,8 +85,7 @@
 ; CHECK-LABEL: test8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleuw (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <32 x i16>, <32 x i16>* %y.ptr, align 4
   %mask = icmp ule <32 x i16> %x, %y
@@ -106,8 +98,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqw %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpcmpeqw %zmm3, %zmm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu16 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp eq <32 x i16> %x1, %y1
   %mask0 = icmp eq <32 x i16> %x, %y
@@ -121,8 +112,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleb %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpcmpleb %zmm2, %zmm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu8 %zmm0, %zmm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %zmm0, %zmm2, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <64 x i8> %x1, %y1
   %mask0 = icmp sle <64 x i8> %x, %y
@@ -136,8 +126,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtb %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vpcmpgtb (%rdi), %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu8 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sgt <64 x i8> %x1, %y1
   %y = load <64 x i8>, <64 x i8>* %y.ptr, align 4
@@ -152,8 +141,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew %zmm1, %zmm2, %k1
 ; CHECK-NEXT:    vpcmpleuw (%rdi), %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu16 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <32 x i16> %x1, %y1
   %y = load <32 x i16>, <32 x i16>* %y.ptr, align 4
Index: llvm/trunk/test/CodeGen/X86/avx512bwvl-mov.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512bwvl-mov.ll
+++ llvm/trunk/test/CodeGen/X86/avx512bwvl-mov.ll
@@ -26,7 +26,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqb %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0x75,0x28,0x3f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu8 (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x07]
+; CHECK-NEXT:    vpblendmb (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x66,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <32 x i8> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <32 x i8>*
@@ -74,7 +74,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqw %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x28,0x3f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu16 (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0xff,0x29,0x6f,0x07]
+; CHECK-NEXT:    vpblendmw (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x66,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <16 x i16> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <16 x i16>*
@@ -122,7 +122,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqb %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x08,0x3f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu8 (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x07]
+; CHECK-NEXT:    vpblendmb (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x66,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <16 x i8> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <16 x i8>*
@@ -170,7 +170,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqw %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x08,0x3f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu16 (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0xff,0x09,0x6f,0x07]
+; CHECK-NEXT:    vpblendmw (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x66,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <8 x i16> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x i16>*
Index: llvm/trunk/test/CodeGen/X86/avx512bwvl-vec-cmp.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512bwvl-vec-cmp.ll
+++ llvm/trunk/test/CodeGen/X86/avx512bwvl-vec-cmp.ll
@@ -5,8 +5,7 @@
 ; CHECK-LABEL: test256_1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k1
-; CHECK-NEXT:    vmovdqu8 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp eq <32 x i8> %x, %y
   %max = select <32 x i1> %mask, <32 x i8> %x, <32 x i8> %y
@@ -17,8 +16,7 @@
 ; CHECK-LABEL: test256_2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k1
-; CHECK-NEXT:    vmovdqu8 %ymm0, %ymm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %ymm0, %ymm2, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sgt <32 x i8> %x, %y
   %max = select <32 x i1> %mask, <32 x i8> %x, <32 x i8> %x1
@@ -29,8 +27,7 @@
 ; CHECK-LABEL: test256_3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew %ymm0, %ymm1, %k1
-; CHECK-NEXT:    vmovdqu16 %ymm2, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm2, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sge <16 x i16> %x, %y
   %max = select <16 x i1> %mask, <16 x i16> %x1, <16 x i16> %y
@@ -41,8 +38,7 @@
 ; CHECK-LABEL: test256_4:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k1
-; CHECK-NEXT:    vmovdqu8 %ymm0, %ymm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %ymm0, %ymm2, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ugt <32 x i8> %x, %y
   %max = select <32 x i1> %mask, <32 x i8> %x, <32 x i8> %x1
@@ -53,8 +49,7 @@
 ; CHECK-LABEL: test256_5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqw (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i16>, <16 x i16>* %yp, align 4
   %mask = icmp eq <16 x i16> %x, %y
@@ -66,8 +61,7 @@
 ; CHECK-LABEL: test256_6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtw (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i16>, <16 x i16>* %y.ptr, align 4
   %mask = icmp sgt <16 x i16> %x, %y
@@ -79,8 +73,7 @@
 ; CHECK-LABEL: test256_7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i16>, <16 x i16>* %y.ptr, align 4
   %mask = icmp sle <16 x i16> %x, %y
@@ -92,8 +85,7 @@
 ; CHECK-LABEL: test256_8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleuw (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <16 x i16>, <16 x i16>* %y.ptr, align 4
   %mask = icmp ule <16 x i16> %x, %y
@@ -106,8 +98,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k1
 ; CHECK-NEXT:    vpcmpeqw %ymm3, %ymm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp eq <16 x i16> %x1, %y1
   %mask0 = icmp eq <16 x i16> %x, %y
@@ -121,8 +112,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k1
 ; CHECK-NEXT:    vpcmpleb %ymm2, %ymm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu8 %ymm0, %ymm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %ymm0, %ymm2, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <32 x i8> %x1, %y1
   %mask0 = icmp sle <32 x i8> %x, %y
@@ -136,8 +126,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtb %ymm2, %ymm1, %k1
 ; CHECK-NEXT:    vpcmpgtb (%rdi), %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu8 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sgt <32 x i8> %x1, %y1
   %y = load <32 x i8>, <32 x i8>* %y.ptr, align 4
@@ -152,8 +141,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew %ymm1, %ymm2, %k1
 ; CHECK-NEXT:    vpcmpleuw (%rdi), %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <16 x i16> %x1, %y1
   %y = load <16 x i16>, <16 x i16>* %y.ptr, align 4
@@ -167,8 +155,7 @@
 ; CHECK-LABEL: test128_1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqu8 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp eq <16 x i8> %x, %y
   %max = select <16 x i1> %mask, <16 x i8> %x, <16 x i8> %y
@@ -179,8 +166,7 @@
 ; CHECK-LABEL: test128_2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqu8 %xmm0, %xmm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %xmm0, %xmm2, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sgt <16 x i8> %x, %y
   %max = select <16 x i1> %mask, <16 x i8> %x, <16 x i8> %x1
@@ -191,8 +177,7 @@
 ; CHECK-LABEL: test128_3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew %xmm0, %xmm1, %k1
-; CHECK-NEXT:    vmovdqu16 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm2, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sge <8 x i16> %x, %y
   %max = select <8 x i1> %mask, <8 x i16> %x1, <8 x i16> %y
@@ -203,8 +188,7 @@
 ; CHECK-LABEL: test128_4:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqu8 %xmm0, %xmm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %xmm0, %xmm2, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ugt <16 x i8> %x, %y
   %max = select <16 x i1> %mask, <16 x i8> %x, <16 x i8> %x1
@@ -215,8 +199,7 @@
 ; CHECK-LABEL: test128_5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqw (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i16>, <8 x i16>* %yp, align 4
   %mask = icmp eq <8 x i16> %x, %y
@@ -228,8 +211,7 @@
 ; CHECK-LABEL: test128_6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtw (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i16>, <8 x i16>* %y.ptr, align 4
   %mask = icmp sgt <8 x i16> %x, %y
@@ -241,8 +223,7 @@
 ; CHECK-LABEL: test128_7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i16>, <8 x i16>* %y.ptr, align 4
   %mask = icmp sle <8 x i16> %x, %y
@@ -254,8 +235,7 @@
 ; CHECK-LABEL: test128_8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleuw (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqu16 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i16>, <8 x i16>* %y.ptr, align 4
   %mask = icmp ule <8 x i16> %x, %y
@@ -268,8 +248,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vpcmpeqw %xmm3, %xmm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu16 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp eq <8 x i16> %x1, %y1
   %mask0 = icmp eq <8 x i16> %x, %y
@@ -283,8 +262,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vpcmpleb %xmm2, %xmm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu8 %xmm0, %xmm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmb %xmm0, %xmm2, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <16 x i8> %x1, %y1
   %mask0 = icmp sle <16 x i8> %x, %y
@@ -298,8 +276,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtb %xmm2, %xmm1, %k1
 ; CHECK-NEXT:    vpcmpgtb (%rdi), %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu8 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sgt <16 x i8> %x1, %y1
   %y = load <16 x i8>, <16 x i8>* %y.ptr, align 4
@@ -314,8 +291,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmplew %xmm1, %xmm2, %k1
 ; CHECK-NEXT:    vpcmpleuw (%rdi), %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqu16 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i16> %x1, %y1
   %y = load <8 x i16>, <8 x i16>* %y.ptr, align 4
Index: llvm/trunk/test/CodeGen/X86/avx512vl-mov.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512vl-mov.ll
+++ llvm/trunk/test/CodeGen/X86/avx512vl-mov.ll
@@ -166,7 +166,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0x75,0x28,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqa32 (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x6f,0x07]
+; CHECK-NEXT:    vpblendmd (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <8 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x i32>*
@@ -180,7 +180,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0x75,0x28,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu32 (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7e,0x29,0x6f,0x07]
+; CHECK-NEXT:    vpblendmd (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <8 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x i32>*
@@ -222,7 +222,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x28,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqa64 (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x29,0x6f,0x07]
+; CHECK-NEXT:    vpblendmq (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x i64>*
@@ -236,7 +236,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x28,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu64 (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0xfe,0x29,0x6f,0x07]
+; CHECK-NEXT:    vpblendmq (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x i64>*
@@ -279,7 +279,7 @@
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vcmpordps %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf1,0x74,0x28,0xc2,0xca,0x07]
 ; CHECK-NEXT:    vcmpneqps %ymm2, %ymm1, %k1 {%k1} ## encoding: [0x62,0xf1,0x74,0x29,0xc2,0xca,0x04]
-; CHECK-NEXT:    vmovaps (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x29,0x28,0x07]
+; CHECK-NEXT:    vblendmps (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x float> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x float>*
@@ -294,7 +294,7 @@
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vcmpordps %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf1,0x74,0x28,0xc2,0xca,0x07]
 ; CHECK-NEXT:    vcmpneqps %ymm2, %ymm1, %k1 {%k1} ## encoding: [0x62,0xf1,0x74,0x29,0xc2,0xca,0x04]
-; CHECK-NEXT:    vmovups (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x29,0x10,0x07]
+; CHECK-NEXT:    vblendmps (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x float> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <8 x float>*
@@ -338,7 +338,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x28,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovapd (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x29,0x28,0x07]
+; CHECK-NEXT:    vblendmpd (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x double>*
@@ -352,7 +352,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %ymm2, %ymm2, %ymm2 ## encoding: [0x62,0xf1,0x6d,0x28,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x28,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovupd (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x29,0x10,0x07]
+; CHECK-NEXT:    vblendmpd (%rdi), %ymm0, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x double>*
@@ -554,7 +554,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqa32 (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x6f,0x07]
+; CHECK-NEXT:    vpblendmd (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x i32>*
@@ -568,7 +568,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu32 (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0x7e,0x09,0x6f,0x07]
+; CHECK-NEXT:    vpblendmd (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x i32>*
@@ -610,7 +610,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqa64 (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x09,0x6f,0x07]
+; CHECK-NEXT:    vpblendmq (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <2 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <2 x i64>*
@@ -624,7 +624,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovdqu64 (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0xfe,0x09,0x6f,0x07]
+; CHECK-NEXT:    vpblendmq (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x64,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <2 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <2 x i64>*
@@ -666,7 +666,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovaps (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x09,0x28,0x07]
+; CHECK-NEXT:    vblendmps (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x float>*
@@ -680,7 +680,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqd %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0x75,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovups (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x09,0x10,0x07]
+; CHECK-NEXT:    vblendmps (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <4 x i32> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <4 x float>*
@@ -722,7 +722,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovapd (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x09,0x28,0x07]
+; CHECK-NEXT:    vblendmpd (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <2 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <2 x double>*
@@ -736,7 +736,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxord %xmm2, %xmm2, %xmm2 ## encoding: [0x62,0xf1,0x6d,0x08,0xef,0xd2]
 ; CHECK-NEXT:    vpcmpneqq %xmm2, %xmm1, %k1 ## encoding: [0x62,0xf3,0xf5,0x08,0x1f,0xca,0x04]
-; CHECK-NEXT:    vmovupd (%rdi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x09,0x10,0x07]
+; CHECK-NEXT:    vblendmpd (%rdi), %xmm0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x65,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = icmp ne <2 x i64> %mask1, zeroinitializer
   %vaddr = bitcast i8* %addr to <2 x double>*
Index: llvm/trunk/test/CodeGen/X86/avx512vl-vec-cmp.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512vl-vec-cmp.ll
+++ llvm/trunk/test/CodeGen/X86/avx512vl-vec-cmp.ll
@@ -5,8 +5,7 @@
 ; CHECK-LABEL: test256_1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp eq <4 x i64> %x, %y
   %max = select <4 x i1> %mask, <4 x i64> %x, <4 x i64> %y
@@ -17,8 +16,7 @@
 ; CHECK-LABEL: test256_2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa64 %ymm2, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm2, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sgt <4 x i64> %x, %y
   %max = select <4 x i1> %mask, <4 x i64> %x1, <4 x i64> %y
@@ -29,8 +27,7 @@
 ; CHECK-LABEL: test256_3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %ymm0, %ymm1, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm2, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm2, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sge <8 x i32> %x, %y
   %max = select <8 x i1> %mask, <8 x i32> %x1, <8 x i32> %y
@@ -41,8 +38,7 @@
 ; CHECK-LABEL: test256_4:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa64 %ymm2, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm2, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ugt <4 x i64> %x, %y
   %max = select <4 x i1> %mask, <4 x i64> %x1, <4 x i64> %y
@@ -53,8 +49,7 @@
 ; CHECK-LABEL: test256_5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %yp, align 4
   %mask = icmp eq <8 x i32> %x, %y
@@ -66,8 +61,7 @@
 ; CHECK-LABEL: test256_6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
   %mask = icmp sgt <8 x i32> %x, %y
@@ -79,8 +73,7 @@
 ; CHECK-LABEL: test256_7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
   %mask = icmp sle <8 x i32> %x, %y
@@ -92,8 +85,7 @@
 ; CHECK-LABEL: test256_8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
   %mask = icmp ule <8 x i32> %x, %y
@@ -106,8 +98,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k1
 ; CHECK-NEXT:    vpcmpeqd %ymm3, %ymm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp eq <8 x i32> %x1, %y1
   %mask0 = icmp eq <8 x i32> %x, %y
@@ -121,8 +112,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k1
 ; CHECK-NEXT:    vpcmpleq %ymm2, %ymm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm0, %ymm2, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <4 x i64> %x1, %y1
   %mask0 = icmp sle <4 x i64> %x, %y
@@ -136,8 +126,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %ymm2, %ymm1, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sgt <4 x i64> %x1, %y1
   %y = load <4 x i64>, <4 x i64>* %y.ptr, align 4
@@ -152,8 +141,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %ymm1, %ymm2, %k1
 ; CHECK-NEXT:    vpcmpleud (%rdi), %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i32> %x1, %y1
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
@@ -167,8 +155,7 @@
 ; CHECK-LABEL: test256_13:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to4}, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %yb = load i64, i64* %yb.ptr, align 4
   %y.0 = insertelement <4 x i64> undef, i64 %yb, i32 0
@@ -182,8 +169,7 @@
 ; CHECK-LABEL: test256_14:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi){1to8}, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %yb = load i32, i32* %yb.ptr, align 4
   %y.0 = insertelement <8 x i32> undef, i32 %yb, i32 0
@@ -198,8 +184,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %ymm1, %ymm2, %k1
 ; CHECK-NEXT:    vpcmpgtd (%rdi){1to8}, %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <8 x i32> %x1, %y1
   %yb = load i32, i32* %yb.ptr, align 4
@@ -216,8 +201,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %ymm1, %ymm2, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi){1to4}, %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <4 x i64> %x1, %y1
   %yb = load i64, i64* %yb.ptr, align 4
@@ -233,8 +217,7 @@
 ; CHECK-LABEL: test128_1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp eq <2 x i64> %x, %y
   %max = select <2 x i1> %mask, <2 x i64> %x, <2 x i64> %y
@@ -245,8 +228,7 @@
 ; CHECK-LABEL: test128_2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm2, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sgt <2 x i64> %x, %y
   %max = select <2 x i1> %mask, <2 x i64> %x1, <2 x i64> %y
@@ -257,8 +239,7 @@
 ; CHECK-LABEL: test128_3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %xmm0, %xmm1, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm2, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp sge <4 x i32> %x, %y
   %max = select <4 x i1> %mask, <4 x i32> %x1, <4 x i32> %y
@@ -269,8 +250,7 @@
 ; CHECK-LABEL: test128_4:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm2, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask = icmp ugt <2 x i64> %x, %y
   %max = select <2 x i1> %mask, <2 x i64> %x1, <2 x i64> %y
@@ -281,8 +261,7 @@
 ; CHECK-LABEL: test128_5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %yp, align 4
   %mask = icmp eq <4 x i32> %x, %y
@@ -294,8 +273,7 @@
 ; CHECK-LABEL: test128_6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp sgt <4 x i32> %x, %y
@@ -307,8 +285,7 @@
 ; CHECK-LABEL: test128_7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp sle <4 x i32> %x, %y
@@ -320,8 +297,7 @@
 ; CHECK-LABEL: test128_8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp ule <4 x i32> %x, %y
@@ -334,8 +310,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vpcmpeqd %xmm3, %xmm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp eq <4 x i32> %x1, %y1
   %mask0 = icmp eq <4 x i32> %x, %y
@@ -349,8 +324,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vpcmpleq %xmm2, %xmm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm2 {%k1}
-; CHECK-NEXT:    vmovaps %zmm2, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm0, %xmm2, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <2 x i64> %x1, %y1
   %mask0 = icmp sle <2 x i64> %x, %y
@@ -364,8 +338,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %xmm2, %xmm1, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sgt <2 x i64> %x1, %y1
   %y = load <2 x i64>, <2 x i64>* %y.ptr, align 4
@@ -380,8 +353,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %xmm1, %xmm2, %k1
 ; CHECK-NEXT:    vpcmpleud (%rdi), %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <4 x i32> %x1, %y1
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
@@ -395,8 +367,7 @@
 ; CHECK-LABEL: test128_13:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to2}, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %yb = load i64, i64* %yb.ptr, align 4
   %y.0 = insertelement <2 x i64> undef, i64 %yb, i32 0
@@ -410,8 +381,7 @@
 ; CHECK-LABEL: test128_14:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi){1to4}, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %yb = load i32, i32* %yb.ptr, align 4
   %y.0 = insertelement <4 x i32> undef, i32 %yb, i32 0
@@ -426,8 +396,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %xmm1, %xmm2, %k1
 ; CHECK-NEXT:    vpcmpgtd (%rdi){1to4}, %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <4 x i32> %x1, %y1
   %yb = load i32, i32* %yb.ptr, align 4
@@ -444,8 +413,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %xmm1, %xmm2, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi){1to2}, %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovaps %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmq %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %mask1 = icmp sge <2 x i64> %x1, %y1
   %yb = load i64, i64* %yb.ptr, align 4
Index: llvm/trunk/test/CodeGen/X86/masked_gather_scatter.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/masked_gather_scatter.ll
+++ llvm/trunk/test/CodeGen/X86/masked_gather_scatter.ll
@@ -1490,8 +1490,7 @@
 ; SKX-NEXT:    vmovq %xmm1, %rax
 ; SKX-NEXT:    vpinsrd $2, (%rax), %xmm0, %xmm0
 ; SKX-NEXT:  .LBB29_6: # %else5
-; SKX-NEXT:    vmovdqa32 %xmm0, %xmm3 {%k1}
-; SKX-NEXT:    vmovaps %zmm3, %zmm0
+; SKX-NEXT:    vpblendmd %xmm0, %xmm3, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test30:
@@ -1503,33 +1502,33 @@
 ; SKX_32-NEXT:    vptestmd %xmm2, %xmm2, %k1
 ; SKX_32-NEXT:    kmovb %k1, {{[0-9]+}}(%esp)
 ; SKX_32-NEXT:    vpslld $2, %xmm1, %xmm1
-; SKX_32-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
+; SKX_32-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
 ; SKX_32-NEXT:    movb {{[0-9]+}}(%esp), %al
-; SKX_32-NEXT:    # implicit-def: %XMM1
+; SKX_32-NEXT:    # implicit-def: %XMM0
 ; SKX_32-NEXT:    testb %al, %al
 ; SKX_32-NEXT:    je .LBB29_2
 ; SKX_32-NEXT:  # BB#1: # %cond.load
-; SKX_32-NEXT:    vmovd %xmm2, %eax
-; SKX_32-NEXT:    vmovd (%eax), %xmm1
+; SKX_32-NEXT:    vmovd %xmm1, %eax
+; SKX_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SKX_32-NEXT:  .LBB29_2: # %else
 ; SKX_32-NEXT:    kmovb %k1, {{[0-9]+}}(%esp)
 ; SKX_32-NEXT:    movb {{[0-9]+}}(%esp), %al
 ; SKX_32-NEXT:    testb %al, %al
 ; SKX_32-NEXT:    je .LBB29_4
 ; SKX_32-NEXT:  # BB#3: # %cond.load1
-; SKX_32-NEXT:    vpextrd $1, %xmm2, %eax
-; SKX_32-NEXT:    vpinsrd $1, (%eax), %xmm1, %xmm1
+; SKX_32-NEXT:    vpextrd $1, %xmm1, %eax
+; SKX_32-NEXT:    vpinsrd $1, (%eax), %xmm0, %xmm0
 ; SKX_32-NEXT:  .LBB29_4: # %else2
-; SKX_32-NEXT:    vmovdqa32 {{[0-9]+}}(%esp), %xmm0
+; SKX_32-NEXT:    vmovdqa32 {{[0-9]+}}(%esp), %xmm2
 ; SKX_32-NEXT:    kmovb %k1, (%esp)
 ; SKX_32-NEXT:    movb (%esp), %al
 ; SKX_32-NEXT:    testb %al, %al
 ; SKX_32-NEXT:    je .LBB29_6
 ; SKX_32-NEXT:  # BB#5: # %cond.load4
-; SKX_32-NEXT:    vpextrd $2, %xmm2, %eax
-; SKX_32-NEXT:    vpinsrd $2, (%eax), %xmm1, %xmm1
+; SKX_32-NEXT:    vpextrd $2, %xmm1, %eax
+; SKX_32-NEXT:    vpinsrd $2, (%eax), %xmm0, %xmm0
 ; SKX_32-NEXT:  .LBB29_6: # %else5
-; SKX_32-NEXT:    vmovdqa32 %xmm1, %xmm0 {%k1}
+; SKX_32-NEXT:    vpblendmd %xmm0, %xmm2, %xmm0 {%k1}
 ; SKX_32-NEXT:    addl $12, %esp
 ; SKX_32-NEXT:    retl