Index: llvm/trunk/lib/Target/X86/X86InstrAVX512.td
===================================================================
--- llvm/trunk/lib/Target/X86/X86InstrAVX512.td
+++ llvm/trunk/lib/Target/X86/X86InstrAVX512.td
@@ -2738,7 +2738,7 @@
                     [(set _.RC:$dst, (_.VT (bitconvert (ld_frag addr:$src))))],
                     _.ExeDomain>, EVEX;
 
-  let Constraints = "$src0 = $dst" in {
+  let Constraints = "$src0 = $dst", isConvertibleToThreeAddress = 1 in {
   def rrk : AVX512PI<opc, MRMSrcReg, (outs _.RC:$dst),
                     (ins _.RC:$src0, _.KRCWM:$mask, _.RC:$src1),
                     !strconcat(OpcodeStr, "\t{$src1, ${dst} {${mask}}|",
Index: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
===================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
@@ -4044,6 +4044,131 @@
         BuildMI(MF, MI.getDebugLoc(), get(X86::LEA16r)).add(Dest).add(Src),
         MI.getOperand(2));
     break;
+
+  case X86::VMOVDQU8Z128rmk:
+  case X86::VMOVDQU8Z256rmk:
+  case X86::VMOVDQU8Zrmk:
+  case X86::VMOVDQU16Z128rmk:
+  case X86::VMOVDQU16Z256rmk:
+  case X86::VMOVDQU16Zrmk:
+  case X86::VMOVDQU32Z128rmk: case X86::VMOVDQA32Z128rmk:
+  case X86::VMOVDQU32Z256rmk: case X86::VMOVDQA32Z256rmk:
+  case X86::VMOVDQU32Zrmk:    case X86::VMOVDQA32Zrmk:
+  case X86::VMOVDQU64Z128rmk: case X86::VMOVDQA64Z128rmk:
+  case X86::VMOVDQU64Z256rmk: case X86::VMOVDQA64Z256rmk:
+  case X86::VMOVDQU64Zrmk:    case X86::VMOVDQA64Zrmk:
+  case X86::VMOVUPDZ128rmk:   case X86::VMOVAPDZ128rmk:
+  case X86::VMOVUPDZ256rmk:   case X86::VMOVAPDZ256rmk:
+  case X86::VMOVUPDZrmk:      case X86::VMOVAPDZrmk:
+  case X86::VMOVUPSZ128rmk:   case X86::VMOVAPSZ128rmk:
+  case X86::VMOVUPSZ256rmk:   case X86::VMOVAPSZ256rmk:
+  case X86::VMOVUPSZrmk:      case X86::VMOVAPSZrmk: {
+    unsigned Opc;
+    switch (MIOpc) {
+    default: llvm_unreachable("Unreachable!");
+    case X86::VMOVDQU8Z128rmk:  Opc = X86::VPBLENDMBZ128rmk; break;
+    case X86::VMOVDQU8Z256rmk:  Opc = X86::VPBLENDMBZ256rmk; break;
+    case X86::VMOVDQU8Zrmk:     Opc = X86::VPBLENDMBZrmk;    break;
+    case X86::VMOVDQU16Z128rmk: Opc = X86::VPBLENDMWZ128rmk; break;
+    case X86::VMOVDQU16Z256rmk: Opc = X86::VPBLENDMWZ256rmk; break;
+    case X86::VMOVDQU16Zrmk:    Opc = X86::VPBLENDMWZrmk;    break;
+    case X86::VMOVDQU32Z128rmk: Opc = X86::VPBLENDMDZ128rmk; break;
+    case X86::VMOVDQU32Z256rmk: Opc = X86::VPBLENDMDZ256rmk; break;
+    case X86::VMOVDQU32Zrmk:    Opc = X86::VPBLENDMDZrmk;    break;
+    case X86::VMOVDQU64Z128rmk: Opc = X86::VPBLENDMQZ128rmk; break;
+    case X86::VMOVDQU64Z256rmk: Opc = X86::VPBLENDMQZ256rmk; break;
+    case X86::VMOVDQU64Zrmk:    Opc = X86::VPBLENDMQZrmk;    break;
+    case X86::VMOVUPDZ128rmk:   Opc = X86::VBLENDMPDZ128rmk; break;
+    case X86::VMOVUPDZ256rmk:   Opc = X86::VBLENDMPDZ256rmk; break;
+    case X86::VMOVUPDZrmk:      Opc = X86::VBLENDMPDZrmk;    break;
+    case X86::VMOVUPSZ128rmk:   Opc = X86::VBLENDMPSZ128rmk; break;
+    case X86::VMOVUPSZ256rmk:   Opc = X86::VBLENDMPSZ256rmk; break;
+    case X86::VMOVUPSZrmk:      Opc = X86::VBLENDMPSZrmk;    break;
+    case X86::VMOVDQA32Z128rmk: Opc = X86::VPBLENDMDZ128rmk; break;
+    case X86::VMOVDQA32Z256rmk: Opc = X86::VPBLENDMDZ256rmk; break;
+    case X86::VMOVDQA32Zrmk:    Opc = X86::VPBLENDMDZrmk;    break;
+    case X86::VMOVDQA64Z128rmk: Opc = X86::VPBLENDMQZ128rmk; break;
+    case X86::VMOVDQA64Z256rmk: Opc = X86::VPBLENDMQZ256rmk; break;
+    case X86::VMOVDQA64Zrmk:    Opc = X86::VPBLENDMQZrmk;    break;
+    case X86::VMOVAPDZ128rmk:   Opc = X86::VBLENDMPDZ128rmk; break;
+    case X86::VMOVAPDZ256rmk:   Opc = X86::VBLENDMPDZ256rmk; break;
+    case X86::VMOVAPDZrmk:      Opc = X86::VBLENDMPDZrmk;    break;
+    case X86::VMOVAPSZ128rmk:   Opc = X86::VBLENDMPSZ128rmk; break;
+    case X86::VMOVAPSZ256rmk:   Opc = X86::VBLENDMPSZ256rmk; break;
+    case X86::VMOVAPSZrmk:      Opc = X86::VBLENDMPSZrmk;    break;
+    }
+
+    NewMI = BuildMI(MF, MI.getDebugLoc(), get(Opc))
+              .add(Dest)
+              .add(MI.getOperand(2))
+              .add(Src)
+              .add(MI.getOperand(3))
+              .add(MI.getOperand(4))
+              .add(MI.getOperand(5))
+              .add(MI.getOperand(6))
+              .add(MI.getOperand(7));
+    break;
+  }
+  case X86::VMOVDQU8Z128rrk:
+  case X86::VMOVDQU8Z256rrk:
+  case X86::VMOVDQU8Zrrk:
+  case X86::VMOVDQU16Z128rrk:
+  case X86::VMOVDQU16Z256rrk:
+  case X86::VMOVDQU16Zrrk:
+  case X86::VMOVDQU32Z128rrk: case X86::VMOVDQA32Z128rrk:
+  case X86::VMOVDQU32Z256rrk: case X86::VMOVDQA32Z256rrk:
+  case X86::VMOVDQU32Zrrk:    case X86::VMOVDQA32Zrrk:
+  case X86::VMOVDQU64Z128rrk: case X86::VMOVDQA64Z128rrk:
+  case X86::VMOVDQU64Z256rrk: case X86::VMOVDQA64Z256rrk:
+  case X86::VMOVDQU64Zrrk:    case X86::VMOVDQA64Zrrk:
+  case X86::VMOVUPDZ128rrk:   case X86::VMOVAPDZ128rrk:
+  case X86::VMOVUPDZ256rrk:   case X86::VMOVAPDZ256rrk:
+  case X86::VMOVUPDZrrk:      case X86::VMOVAPDZrrk:
+  case X86::VMOVUPSZ128rrk:   case X86::VMOVAPSZ128rrk:
+  case X86::VMOVUPSZ256rrk:   case X86::VMOVAPSZ256rrk:
+  case X86::VMOVUPSZrrk:      case X86::VMOVAPSZrrk: {
+    unsigned Opc;
+    switch (MIOpc) {
+    default: llvm_unreachable("Unreachable!");
+    case X86::VMOVDQU8Z128rrk:  Opc = X86::VPBLENDMBZ128rrk; break;
+    case X86::VMOVDQU8Z256rrk:  Opc = X86::VPBLENDMBZ256rrk; break;
+    case X86::VMOVDQU8Zrrk:     Opc = X86::VPBLENDMBZrrk;    break;
+    case X86::VMOVDQU16Z128rrk: Opc = X86::VPBLENDMWZ128rrk; break;
+    case X86::VMOVDQU16Z256rrk: Opc = X86::VPBLENDMWZ256rrk; break;
+    case X86::VMOVDQU16Zrrk:    Opc = X86::VPBLENDMWZrrk;    break;
+    case X86::VMOVDQU32Z128rrk: Opc = X86::VPBLENDMDZ128rrk; break;
+    case X86::VMOVDQU32Z256rrk: Opc = X86::VPBLENDMDZ256rrk; break;
+    case X86::VMOVDQU32Zrrk:    Opc = X86::VPBLENDMDZrrk;    break;
+    case X86::VMOVDQU64Z128rrk: Opc = X86::VPBLENDMQZ128rrk; break;
+    case X86::VMOVDQU64Z256rrk: Opc = X86::VPBLENDMQZ256rrk; break;
+    case X86::VMOVDQU64Zrrk:    Opc = X86::VPBLENDMQZrrk;    break;
+    case X86::VMOVUPDZ128rrk:   Opc = X86::VBLENDMPDZ128rrk; break;
+    case X86::VMOVUPDZ256rrk:   Opc = X86::VBLENDMPDZ256rrk; break;
+    case X86::VMOVUPDZrrk:      Opc = X86::VBLENDMPDZrrk;    break;
+    case X86::VMOVUPSZ128rrk:   Opc = X86::VBLENDMPSZ128rrk; break;
+    case X86::VMOVUPSZ256rrk:   Opc = X86::VBLENDMPSZ256rrk; break;
+    case X86::VMOVUPSZrrk:      Opc = X86::VBLENDMPSZrrk;    break;
+    case X86::VMOVDQA32Z128rrk: Opc = X86::VPBLENDMDZ128rrk; break;
+    case X86::VMOVDQA32Z256rrk: Opc = X86::VPBLENDMDZ256rrk; break;
+    case X86::VMOVDQA32Zrrk:    Opc = X86::VPBLENDMDZrrk;    break;
+    case X86::VMOVDQA64Z128rrk: Opc = X86::VPBLENDMQZ128rrk; break;
+    case X86::VMOVDQA64Z256rrk: Opc = X86::VPBLENDMQZ256rrk; break;
+    case X86::VMOVDQA64Zrrk:    Opc = X86::VPBLENDMQZrrk;    break;
+    case X86::VMOVAPDZ128rrk:   Opc = X86::VBLENDMPDZ128rrk; break;
+    case X86::VMOVAPDZ256rrk:   Opc = X86::VBLENDMPDZ256rrk; break;
+    case X86::VMOVAPDZrrk:      Opc = X86::VBLENDMPDZrrk;    break;
+    case X86::VMOVAPSZ128rrk:   Opc = X86::VBLENDMPSZ128rrk; break;
+    case X86::VMOVAPSZ256rrk:   Opc = X86::VBLENDMPSZ256rrk; break;
+    case X86::VMOVAPSZrrk:      Opc = X86::VBLENDMPSZrrk;    break;
+    }
+
+    NewMI = BuildMI(MF, MI.getDebugLoc(), get(Opc))
+              .add(Dest)
+              .add(MI.getOperand(2))
+              .add(Src)
+              .add(MI.getOperand(3));
+    break;
+  }
   }
 
   if (!NewMI) return nullptr;
Index: llvm/trunk/test/CodeGen/X86/avx512-bugfix-26264.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-bugfix-26264.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-bugfix-26264.ll
@@ -6,17 +6,14 @@
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsllw $7, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpmovb2m %zmm0, %k1
-; AVX512BW-NEXT:    vmovupd (%rdi), %zmm1 {%k1}
+; AVX512BW-NEXT:    vblendmpd (%rdi), %zmm1, %zmm0 {%k1}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k2
-; AVX512BW-NEXT:    vmovupd 128(%rdi), %zmm3 {%k2}
+; AVX512BW-NEXT:    vblendmpd 128(%rdi), %zmm3, %zmm5 {%k2}
 ; AVX512BW-NEXT:    kshiftrw $8, %k1, %k1
-; AVX512BW-NEXT:    vmovupd 64(%rdi), %zmm2 {%k1}
+; AVX512BW-NEXT:    vblendmpd 64(%rdi), %zmm2, %zmm1 {%k1}
 ; AVX512BW-NEXT:    kshiftrw $8, %k2, %k1
-; AVX512BW-NEXT:    vmovupd 192(%rdi), %zmm4 {%k1}
-; AVX512BW-NEXT:    vmovapd %zmm1, %zmm0
-; AVX512BW-NEXT:    vmovapd %zmm2, %zmm1
-; AVX512BW-NEXT:    vmovapd %zmm3, %zmm2
-; AVX512BW-NEXT:    vmovapd %zmm4, %zmm3
+; AVX512BW-NEXT:    vblendmpd 192(%rdi), %zmm4, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovapd %zmm5, %zmm2
 ; AVX512BW-NEXT:    retq
   %res = call <32 x double> @llvm.masked.load.v32f64.p0v32f64(<32 x double>* %ptrs, i32 4, <32 x i1> %mask, <32 x double> %src0)
   ret <32 x double> %res
@@ -27,17 +24,14 @@
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpsllw $7, %ymm0, %ymm0
 ; AVX512BW-NEXT:    vpmovb2m %zmm0, %k1
-; AVX512BW-NEXT:    vmovdqu64 (%rdi), %zmm1 {%k1}
+; AVX512BW-NEXT:    vpblendmq (%rdi), %zmm1, %zmm0 {%k1}
 ; AVX512BW-NEXT:    kshiftrd $16, %k1, %k2
-; AVX512BW-NEXT:    vmovdqu64 128(%rdi), %zmm3 {%k2}
+; AVX512BW-NEXT:    vpblendmq 128(%rdi), %zmm3, %zmm5 {%k2}
 ; AVX512BW-NEXT:    kshiftrw $8, %k1, %k1
-; AVX512BW-NEXT:    vmovdqu64 64(%rdi), %zmm2 {%k1}
+; AVX512BW-NEXT:    vpblendmq 64(%rdi), %zmm2, %zmm1 {%k1}
 ; AVX512BW-NEXT:    kshiftrw $8, %k2, %k1
-; AVX512BW-NEXT:    vmovdqu64 192(%rdi), %zmm4 {%k1}
-; AVX512BW-NEXT:    vmovdqa64 %zmm1, %zmm0
-; AVX512BW-NEXT:    vmovdqa64 %zmm2, %zmm1
-; AVX512BW-NEXT:    vmovdqa64 %zmm3, %zmm2
-; AVX512BW-NEXT:    vmovdqa64 %zmm4, %zmm3
+; AVX512BW-NEXT:    vpblendmq 192(%rdi), %zmm4, %zmm3 {%k1}
+; AVX512BW-NEXT:    vmovdqa64 %zmm5, %zmm2
 ; AVX512BW-NEXT:    retq
   %res = call <32 x i64> @llvm.masked.load.v32i64.p0v32i64(<32 x i64>* %ptrs, i32 4, <32 x i1> %mask, <32 x i64> %src0)
   ret <32 x i64> %res
Index: llvm/trunk/test/CodeGen/X86/avx512-masked-memop-64-32.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-masked-memop-64-32.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-masked-memop-64-32.ll
@@ -43,8 +43,7 @@
 ; AVX512:       ## BB#0:
 ; AVX512-NEXT:    vpxord %zmm2, %zmm2, %zmm2
 ; AVX512-NEXT:    vpcmpeqd %zmm2, %zmm0, %k1
-; AVX512-NEXT:    vmovups (%rdi), %zmm1 {%k1}
-; AVX512-NEXT:    vmovaps %zmm1, %zmm0
+; AVX512-NEXT:    vblendmps (%rdi), %zmm1, %zmm0 {%k1}
 ; AVX512-NEXT:    retq
   %mask = icmp eq <16 x i32> %trigger, zeroinitializer
   %res = call <16 x float> @llvm.masked.load.v16f32.p0v16f32(<16 x float>* %addr, i32 4, <16 x i1>%mask, <16 x float> %dst)
@@ -189,22 +188,18 @@
 ; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
 ; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k1
-; AVX512F-NEXT:    vmovdqu64 (%rdi), %zmm1 {%k1}
+; AVX512F-NEXT:    vpblendmq (%rdi), %zmm1, %zmm0 {%k1}
 ; AVX512F-NEXT:    kshiftrw $8, %k1, %k1
-; AVX512F-NEXT:    vmovdqu64 64(%rdi), %zmm2 {%k1}
-; AVX512F-NEXT:    vmovdqa64 %zmm1, %zmm0
-; AVX512F-NEXT:    vmovdqa64 %zmm2, %zmm1
+; AVX512F-NEXT:    vpblendmq 64(%rdi), %zmm2, %zmm1 {%k1}
 ; AVX512F-NEXT:    retq
 ;
 ; SKX-LABEL: test_load_16i64:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpsllw $7, %xmm0, %xmm0
 ; SKX-NEXT:    vpmovb2m %xmm0, %k1
-; SKX-NEXT:    vmovdqu64 (%rdi), %zmm1 {%k1}
+; SKX-NEXT:    vpblendmq (%rdi), %zmm1, %zmm0 {%k1}
 ; SKX-NEXT:    kshiftrw $8, %k1, %k1
-; SKX-NEXT:    vmovdqu64 64(%rdi), %zmm2 {%k1}
-; SKX-NEXT:    vmovdqa64 %zmm1, %zmm0
-; SKX-NEXT:    vmovdqa64 %zmm2, %zmm1
+; SKX-NEXT:    vpblendmq 64(%rdi), %zmm2, %zmm1 {%k1}
 ; SKX-NEXT:    retq
   %res = call <16 x i64> @llvm.masked.load.v16i64.p0v16i64(<16 x i64>* %ptrs, i32 4, <16 x i1> %mask, <16 x i64> %src0)
   ret <16 x i64> %res
@@ -217,22 +212,18 @@
 ; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
 ; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k1
-; AVX512F-NEXT:    vmovupd (%rdi), %zmm1 {%k1}
+; AVX512F-NEXT:    vblendmpd (%rdi), %zmm1, %zmm0 {%k1}
 ; AVX512F-NEXT:    kshiftrw $8, %k1, %k1
-; AVX512F-NEXT:    vmovupd 64(%rdi), %zmm2 {%k1}
-; AVX512F-NEXT:    vmovapd %zmm1, %zmm0
-; AVX512F-NEXT:    vmovapd %zmm2, %zmm1
+; AVX512F-NEXT:    vblendmpd 64(%rdi), %zmm2, %zmm1 {%k1}
 ; AVX512F-NEXT:    retq
 ;
 ; SKX-LABEL: test_load_16f64:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpsllw $7, %xmm0, %xmm0
 ; SKX-NEXT:    vpmovb2m %xmm0, %k1
-; SKX-NEXT:    vmovupd (%rdi), %zmm1 {%k1}
+; SKX-NEXT:    vblendmpd (%rdi), %zmm1, %zmm0 {%k1}
 ; SKX-NEXT:    kshiftrw $8, %k1, %k1
-; SKX-NEXT:    vmovupd 64(%rdi), %zmm2 {%k1}
-; SKX-NEXT:    vmovapd %zmm1, %zmm0
-; SKX-NEXT:    vmovapd %zmm2, %zmm1
+; SKX-NEXT:    vblendmpd 64(%rdi), %zmm2, %zmm1 {%k1}
 ; SKX-NEXT:    retq
   %res = call <16 x double> @llvm.masked.load.v16f64.p0v16f64(<16 x double>* %ptrs, i32 4, <16 x i1> %mask, <16 x double> %src0)
   ret <16 x double> %res
@@ -246,36 +237,30 @@
 ; AVX512F-NEXT:    vpmovsxbd %xmm5, %zmm5
 ; AVX512F-NEXT:    vpslld $31, %zmm5, %zmm5
 ; AVX512F-NEXT:    vptestmd %zmm5, %zmm5, %k1
-; AVX512F-NEXT:    vmovupd 128(%rdi), %zmm3 {%k1}
+; AVX512F-NEXT:    vblendmpd 128(%rdi), %zmm3, %zmm5 {%k1}
 ; AVX512F-NEXT:    vpmovsxbd %xmm0, %zmm0
 ; AVX512F-NEXT:    vpslld $31, %zmm0, %zmm0
 ; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k2
-; AVX512F-NEXT:    vmovupd (%rdi), %zmm1 {%k2}
+; AVX512F-NEXT:    vblendmpd (%rdi), %zmm1, %zmm0 {%k2}
 ; AVX512F-NEXT:    kshiftrw $8, %k1, %k1
-; AVX512F-NEXT:    vmovupd 192(%rdi), %zmm4 {%k1}
+; AVX512F-NEXT:    vblendmpd 192(%rdi), %zmm4, %zmm3 {%k1}
 ; AVX512F-NEXT:    kshiftrw $8, %k2, %k1
-; AVX512F-NEXT:    vmovupd 64(%rdi), %zmm2 {%k1}
-; AVX512F-NEXT:    vmovapd %zmm1, %zmm0
-; AVX512F-NEXT:    vmovapd %zmm2, %zmm1
-; AVX512F-NEXT:    vmovapd %zmm3, %zmm2
-; AVX512F-NEXT:    vmovapd %zmm4, %zmm3
+; AVX512F-NEXT:    vblendmpd 64(%rdi), %zmm2, %zmm1 {%k1}
+; AVX512F-NEXT:    vmovapd %zmm5, %zmm2
 ; AVX512F-NEXT:    retq
 ;
 ; SKX-LABEL: test_load_32f64:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpsllw $7, %ymm0, %ymm0
 ; SKX-NEXT:    vpmovb2m %ymm0, %k1
-; SKX-NEXT:    vmovupd (%rdi), %zmm1 {%k1}
+; SKX-NEXT:    vblendmpd (%rdi), %zmm1, %zmm0 {%k1}
 ; SKX-NEXT:    kshiftrd $16, %k1, %k2
-; SKX-NEXT:    vmovupd 128(%rdi), %zmm3 {%k2}
+; SKX-NEXT:    vblendmpd 128(%rdi), %zmm3, %zmm5 {%k2}
 ; SKX-NEXT:    kshiftrw $8, %k1, %k1
-; SKX-NEXT:    vmovupd 64(%rdi), %zmm2 {%k1}
+; SKX-NEXT:    vblendmpd 64(%rdi), %zmm2, %zmm1 {%k1}
 ; SKX-NEXT:    kshiftrw $8, %k2, %k1
-; SKX-NEXT:    vmovupd 192(%rdi), %zmm4 {%k1}
-; SKX-NEXT:    vmovapd %zmm1, %zmm0
-; SKX-NEXT:    vmovapd %zmm2, %zmm1
-; SKX-NEXT:    vmovapd %zmm3, %zmm2
-; SKX-NEXT:    vmovapd %zmm4, %zmm3
+; SKX-NEXT:    vblendmpd 192(%rdi), %zmm4, %zmm3 {%k1}
+; SKX-NEXT:    vmovapd %zmm5, %zmm2
 ; SKX-NEXT:    retq
   %res = call <32 x double> @llvm.masked.load.v32f64.p0v32f64(<32 x double>* %ptrs, i32 4, <32 x i1> %mask, <32 x double> %src0)
   ret <32 x double> %res
Index: llvm/trunk/test/CodeGen/X86/avx512-masked_memop-16-8.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-masked_memop-16-8.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-masked_memop-16-8.ll
@@ -20,8 +20,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpsllw $7, %ymm0, %ymm0
 ; CHECK-NEXT:    vpmovb2m %ymm0, %k1
-; CHECK-NEXT:    vmovdqu8 (%rdi), %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmb (%rdi), %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %res = call <32 x i8> @llvm.masked.load.v32i8.p0v32i8(<32 x i8>* %addr, i32 4, <32 x i1>%mask, <32 x i8> %val)
   ret <32 x i8> %res
@@ -33,8 +32,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpsllw $7, %zmm0, %zmm0
 ; CHECK-NEXT:    vpmovb2m %zmm0, %k1
-; CHECK-NEXT:    vmovdqu8 (%rdi), %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmb (%rdi), %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %res = call <64 x i8> @llvm.masked.load.v64i8.p0v64i8(<64 x i8>* %addr, i32 4, <64 x i1>%mask, <64 x i8> %val)
   ret <64 x i8> %res
@@ -70,8 +68,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpsllw $7, %ymm0, %ymm0
 ; CHECK-NEXT:    vpmovb2m %ymm0, %k1
-; CHECK-NEXT:    vmovdqu16 (%rdi), %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    vpblendmw (%rdi), %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
   %res = call <32 x i16> @llvm.masked.load.v32i16.p0v32i16(<32 x i16>* %addr, i32 4, <32 x i1>%mask, <32 x i16> %val)
   ret <32 x i16> %res
Index: llvm/trunk/test/CodeGen/X86/avx512-regcall-NoMask.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-regcall-NoMask.ll
+++ llvm/trunk/test/CodeGen/X86/avx512-regcall-NoMask.ll
@@ -325,13 +325,11 @@
 }
 
 ; X32-LABEL:  test_argRet128Vector:
-; X32:        vmovdqa{{.*}}  %xmm0, %xmm1
-; X32:        vmovdqa{{.*}}  %xmm1, %xmm0
+; X32:        vpblend{{.*}}  %xmm0, %xmm1, %xmm0
 ; X32:        ret{{.*}}
 
 ; WIN64-LABEL:  test_argRet128Vector:
-; WIN64:        vmovdqa{{.*}}  %xmm0, %xmm1
-; WIN64:        vmovdqa{{.*}}  %xmm1, %xmm0
+; WIN64:        vpblend{{.*}}  %xmm0, %xmm1, %xmm0
 ; WIN64:        ret{{.*}}
 
 ; Test regcall when receiving/returning 128 bit vector
@@ -360,13 +358,11 @@
 }
 
 ; X32-LABEL:  test_argRet256Vector:
-; X32:        vmovdqa{{.*}}  %ymm0, %ymm1
-; X32:        vmovdqa{{.*}}  %ymm1, %ymm0
+; X32:        vpblend{{.*}}  %ymm0, %ymm1, %ymm0
 ; X32:        ret{{.*}}
 
 ; WIN64-LABEL:  test_argRet256Vector:
-; WIN64:        vmovdqa{{.*}}  %ymm0, %ymm1
-; WIN64:        vmovdqa{{.*}}  %ymm1, %ymm0
+; WIN64:        vpblend{{.*}}  %ymm0, %ymm1, %ymm0
 ; WIN64:        ret{{.*}}
 
 ; Test regcall when receiving/returning 256 bit vector
@@ -395,13 +391,11 @@
 }
 
 ; X32-LABEL:  test_argRet512Vector:
-; X32:        vmovdqa{{.*}}  %zmm0, %zmm1
-; X32:        vmovdqa{{.*}}  %zmm1, %zmm0
+; X32:        vpblend{{.*}}  %zmm0, %zmm1, %zmm0
 ; X32:        ret{{.*}}
 
 ; WIN64-LABEL:  test_argRet512Vector:
-; WIN64:        vmovdqa{{.*}}  %zmm0, %zmm1
-; WIN64:        vmovdqa{{.*}}  %zmm1, %zmm0
+; WIN64:        vpblend{{.*}}  %zmm0, %zmm1, %zmm0
 ; WIN64:        ret{{.*}}
 
 ; Test regcall when receiving/returning 512 bit vector
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:    vmovapd %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vxorps %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vcmpltps %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovaps %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %xmm1, %xmm0
+; 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:    vxorpd %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vcmpltpd %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovapd %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovapd %xmm1, %xmm0
+; 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
@@ -122,15 +114,14 @@
 ; KNL-NEXT:    ## kill: %YMM1<def> %YMM1<kill> %ZMM1<def>
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<def>
 ; KNL-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
-; KNL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vmovdqa %ymm1, %ymm0
+; KNL-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
+; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test9:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpcmpeqd %ymm1, %ymm0, %k1
-; SKX-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -143,15 +134,14 @@
 ; KNL-NEXT:    ## kill: %YMM1<def> %YMM1<kill> %ZMM1<def>
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<def>
 ; KNL-NEXT:    vcmpeqps %zmm1, %zmm0, %k1
-; KNL-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vmovaps %ymm1, %ymm0
+; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
+; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test10:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpeqps %ymm1, %ymm0, %k1
-; SKX-NEXT:    vmovaps %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %ymm1, %ymm0
+; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %mask = fcmp oeq <8 x float> %x, %y
@@ -699,8 +689,7 @@
 ; CHECK-LABEL: test16:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm2, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -711,8 +700,7 @@
 ; CHECK-LABEL: test17:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -724,8 +712,7 @@
 ; CHECK-LABEL: test18:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -737,8 +724,7 @@
 ; CHECK-LABEL: test19:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -751,8 +737,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpcmpeqd %zmm3, %zmm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -766,8 +751,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vpcmpleq %zmm2, %zmm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm2 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -781,8 +765,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -797,8 +780,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %zmm1, %zmm2, %k1
 ; CHECK-NEXT:    vpcmpleud (%rdi), %zmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -812,8 +794,7 @@
 ; CHECK-LABEL: test24:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to8}, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -827,8 +808,7 @@
 ; CHECK-LABEL: test25:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi){1to16}, %zmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %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
@@ -843,8 +823,7 @@
 ; 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:    vmovdqa64 %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
@@ -861,8 +840,7 @@
 ; 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:    vmovdqa64 %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
@@ -932,8 +910,7 @@
 ; SKX-LABEL: test30:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpeqpd %ymm1, %ymm0, %k1
-; SKX-NEXT:    vmovapd %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovapd %ymm1, %ymm0
+; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %mask = fcmp oeq <4 x double> %x, %y
@@ -951,8 +928,7 @@
 ; SKX-LABEL: test31:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi), %xmm0, %k1
-; SKX-NEXT:    vmovapd %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovapd %xmm1, %xmm0
+; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %y = load <2 x double>, <2 x double>* %yp, align 4
@@ -971,8 +947,7 @@
 ; SKX-LABEL: test32:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi), %ymm0, %k1
-; SKX-NEXT:    vmovapd %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovapd %ymm1, %ymm0
+; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %y = load <4 x double>, <4 x double>* %yp, align 4
@@ -985,8 +960,7 @@
 ; CHECK-LABEL: test33:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpltpd (%rdi), %zmm0, %k1
-; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovapd %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
@@ -1004,8 +978,7 @@
 ; SKX-LABEL: test34:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi), %xmm0, %k1
-; SKX-NEXT:    vmovaps %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %xmm1, %xmm0
+; 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
@@ -1020,15 +993,14 @@
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<def>
 ; KNL-NEXT:    vmovups (%rdi), %ymm2
 ; KNL-NEXT:    vcmpltps %zmm2, %zmm0, %k1
-; KNL-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vmovaps %ymm1, %ymm0
+; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
+; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test35:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi), %ymm0, %k1
-; SKX-NEXT:    vmovaps %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %ymm1, %ymm0
+; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %y = load <8 x float>, <8 x float>* %yp, align 4
@@ -1041,8 +1013,7 @@
 ; 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
@@ -1054,8 +1025,7 @@
 ; CHECK-LABEL: test37:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpltpd (%rdi){1to8}, %zmm0, %k1
-; CHECK-NEXT:    vmovapd %zmm0, %zmm1 {%k1}
-; CHECK-NEXT:    vmovapd %zmm1, %zmm0
+; CHECK-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; CHECK-NEXT:    retq
 
   %a = load double, double* %ptr
@@ -1078,8 +1048,7 @@
 ; SKX-LABEL: test38:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi){1to4}, %ymm0, %k1
-; SKX-NEXT:    vmovapd %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovapd %ymm1, %ymm0
+; SKX-NEXT:    vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load double, double* %ptr
@@ -1102,8 +1071,7 @@
 ; SKX-LABEL: test39:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltpd (%rdi){1to2}, %xmm0, %k1
-; SKX-NEXT:    vmovapd %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovapd %xmm1, %xmm0
+; SKX-NEXT:    vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load double, double* %ptr
@@ -1120,8 +1088,7 @@
 ; 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
@@ -1140,15 +1107,14 @@
 ; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<def>
 ; KNL-NEXT:    vbroadcastss (%rdi), %ymm2
 ; KNL-NEXT:    vcmpltps %zmm2, %zmm0, %k1
-; KNL-NEXT:    vmovaps %zmm0, %zmm1 {%k1}
-; KNL-NEXT:    vmovaps %ymm1, %ymm0
+; KNL-NEXT:    vblendmps %zmm0, %zmm1, %zmm0 {%k1}
+; KNL-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test41:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi){1to8}, %ymm0, %k1
-; SKX-NEXT:    vmovaps %ymm0, %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %ymm1, %ymm0
+; SKX-NEXT:    vblendmps %ymm0, %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load float, float* %ptr
@@ -1171,8 +1137,7 @@
 ; SKX-LABEL: test42:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vcmpltps (%rdi){1to4}, %xmm0, %k1
-; SKX-NEXT:    vmovaps %xmm0, %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %xmm1, %xmm0
+; SKX-NEXT:    vblendmps %xmm0, %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load float, float* %ptr
@@ -1191,8 +1156,7 @@
 ; 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:    vmovapd %zmm1, %zmm0
+; KNL-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: test43:
@@ -1200,8 +1164,7 @@
 ; 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:    vmovapd %zmm1, %zmm0
+; SKX-NEXT:    vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
 ; SKX-NEXT:    retq
 
   %a = load double, double* %ptr
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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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:    vmovdqa64 %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-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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm2, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm2, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm2, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm2, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm2, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm2, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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:    vmovdqa %xmm1, %xmm0
+; 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-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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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:    vmovdqa %ymm1, %ymm0
+; 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_5b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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> %y, %x
@@ -79,8 +73,7 @@
 ; CHECK-LABEL: test256_6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -92,8 +85,7 @@
 ; CHECK-LABEL: test256_6b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
   %mask = icmp slt <8 x i32> %y, %x
@@ -105,8 +97,7 @@
 ; CHECK-LABEL: test256_7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -118,8 +109,7 @@
 ; CHECK-LABEL: test256_7b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
   %mask = icmp sge <8 x i32> %y, %x
@@ -131,8 +121,7 @@
 ; CHECK-LABEL: test256_8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -144,8 +133,7 @@
 ; CHECK-LABEL: test256_8b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %y.ptr, align 4
   %mask = icmp uge <8 x i32> %y, %x
@@ -158,8 +146,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k1
 ; CHECK-NEXT:    vpcmpeqd %ymm3, %ymm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -173,8 +160,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k1
 ; CHECK-NEXT:    vpcmpleq %ymm2, %ymm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm2 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm2, %ymm0
+; 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
@@ -188,8 +174,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %ymm2, %ymm1, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -204,8 +189,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %ymm1, %ymm2, %k1
 ; CHECK-NEXT:    vpcmpleud (%rdi), %ymm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -219,8 +203,7 @@
 ; CHECK-LABEL: test256_13:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to4}, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa64 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -234,8 +217,7 @@
 ; CHECK-LABEL: test256_14:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi){1to8}, %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; 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
@@ -250,8 +232,7 @@
 ; 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:    vmovdqa %ymm1, %ymm0
+; 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
@@ -268,8 +249,7 @@
 ; 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:    vmovdqa %ymm1, %ymm0
+; 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
@@ -285,8 +265,7 @@
 ; CHECK-LABEL: test256_17:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpneqd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %yp, align 4
   %mask = icmp ne <8 x i32> %x, %y
@@ -298,8 +277,7 @@
 ; CHECK-LABEL: test256_18:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpneqd (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %yp, align 4
   %mask = icmp ne <8 x i32> %y, %x
@@ -311,8 +289,7 @@
 ; CHECK-LABEL: test256_19:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnltud (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %yp, align 4
   %mask = icmp uge <8 x i32> %x, %y
@@ -324,8 +301,7 @@
 ; CHECK-LABEL: test256_20:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %ymm0, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <8 x i32>, <8 x i32>* %yp, align 4
   %mask = icmp uge <8 x i32> %y, %x
@@ -337,8 +313,7 @@
 ; CHECK-LABEL: test128_1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -349,8 +324,7 @@
 ; CHECK-LABEL: test128_2:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -361,8 +335,7 @@
 ; CHECK-LABEL: test128_3:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %xmm0, %xmm1, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -373,8 +346,7 @@
 ; CHECK-LABEL: test128_4:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm2, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -385,8 +357,7 @@
 ; CHECK-LABEL: test128_5:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -398,8 +369,7 @@
 ; CHECK-LABEL: test128_5b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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> %y, %x
@@ -411,8 +381,7 @@
 ; CHECK-LABEL: test128_6:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -424,8 +393,7 @@
 ; CHECK-LABEL: test128_6b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp slt <4 x i32> %y, %x
@@ -437,8 +405,7 @@
 ; CHECK-LABEL: test128_7:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -450,8 +417,7 @@
 ; CHECK-LABEL: test128_7b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp sge <4 x i32> %y, %x
@@ -463,8 +429,7 @@
 ; CHECK-LABEL: test128_8:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -476,8 +441,7 @@
 ; CHECK-LABEL: test128_8b:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp uge <4 x i32> %y, %x
@@ -490,8 +454,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vpcmpeqd %xmm3, %xmm2, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -505,8 +468,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vpcmpleq %xmm2, %xmm3, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm2 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm2, %xmm0
+; 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
@@ -520,8 +482,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpgtq %xmm2, %xmm1, %k1
 ; CHECK-NEXT:    vpcmpgtq (%rdi), %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -536,8 +497,7 @@
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled %xmm1, %xmm2, %k1
 ; CHECK-NEXT:    vpcmpleud (%rdi), %xmm0, %k1 {%k1}
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -551,8 +511,7 @@
 ; CHECK-LABEL: test128_13:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpeqq (%rdi){1to2}, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa64 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -566,8 +525,7 @@
 ; CHECK-LABEL: test128_14:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpled (%rdi){1to4}, %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; 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
@@ -582,8 +540,7 @@
 ; 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:    vmovdqa %xmm1, %xmm0
+; 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
@@ -600,8 +557,7 @@
 ; 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:    vmovdqa %xmm1, %xmm0
+; 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
@@ -617,8 +573,7 @@
 ; CHECK-LABEL: test128_17:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpneqd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp ne <4 x i32> %x, %y
@@ -630,8 +585,7 @@
 ; CHECK-LABEL: test128_18:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpneqd (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp ne <4 x i32> %y, %x
@@ -643,8 +597,7 @@
 ; CHECK-LABEL: test128_19:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpnltud (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp uge <4 x i32> %x, %y
@@ -656,8 +609,7 @@
 ; CHECK-LABEL: test128_20:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpcmpleud (%rdi), %xmm0, %k1
-; CHECK-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1}
-; CHECK-NEXT:    vmovdqa %xmm1, %xmm0
+; CHECK-NEXT:    vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
 ; CHECK-NEXT:    retq
   %y = load <4 x i32>, <4 x i32>* %y.ptr, align 4
   %mask = icmp uge <4 x i32> %y, %x
Index: llvm/trunk/test/CodeGen/X86/masked_memop.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/masked_memop.ll
+++ llvm/trunk/test/CodeGen/X86/masked_memop.ll
@@ -29,8 +29,7 @@
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vpcmpeqq %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovupd (%rdi), %xmm1 {%k1}
-; SKX-NEXT:    vmovapd %xmm1, %xmm0
+; SKX-NEXT:    vblendmpd (%rdi), %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <2 x i64> %trigger, zeroinitializer
   %res = call <2 x double> @llvm.masked.load.v2f64.p0v2f64(<2 x double>* %addr, i32 4, <2 x i1>%mask, <2 x double>%dst)
@@ -58,8 +57,7 @@
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vpcmpeqd %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovups (%rdi), %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %xmm1, %xmm0
+; SKX-NEXT:    vblendmps (%rdi), %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <4 x i32> %trigger, zeroinitializer
   %res = call <4 x float> @llvm.masked.load.v4f32.p0v4f32(<4 x float>* %addr, i32 4, <4 x i1>%mask, <4 x float>%dst)
@@ -95,8 +93,7 @@
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vpcmpeqd %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovdqu32 (%rdi), %xmm1 {%k1}
-; SKX-NEXT:    vmovdqa %xmm1, %xmm0
+; SKX-NEXT:    vpblendmd (%rdi), %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <4 x i32> %trigger, zeroinitializer
   %res = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %addr, i32 4, <4 x i1>%mask, <4 x i32>%dst)
@@ -171,8 +168,7 @@
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    vpcmpeqd %xmm2, %xmm0, %k1
-; SKX-NEXT:    vmovapd (%rdi), %ymm1 {%k1}
-; SKX-NEXT:    vmovapd %ymm1, %ymm0
+; SKX-NEXT:    vblendmpd (%rdi), %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <4 x i32> %trigger, zeroinitializer
   %res = call <4 x double> @llvm.masked.load.v4f64.p0v4f64(<4 x double>* %addr, i32 32, <4 x i1>%mask, <4 x double>%dst)
@@ -246,16 +242,15 @@
 ; AVX512F-NEXT:    vpcmpeqd %zmm2, %zmm0, %k0
 ; AVX512F-NEXT:    kshiftlw $8, %k0, %k0
 ; AVX512F-NEXT:    kshiftrw $8, %k0, %k1
-; AVX512F-NEXT:    vmovups (%rdi), %zmm1 {%k1}
-; AVX512F-NEXT:    vmovaps %ymm1, %ymm0
+; AVX512F-NEXT:    vblendmps (%rdi), %zmm1, %zmm0 {%k1}
+; AVX512F-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; AVX512F-NEXT:    retq
 ;
 ; SKX-LABEL: test11a:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpxor %ymm2, %ymm2, %ymm2
 ; SKX-NEXT:    vpcmpeqd %ymm2, %ymm0, %k1
-; SKX-NEXT:    vmovaps (%rdi), %ymm1 {%k1}
-; SKX-NEXT:    vmovaps %ymm1, %ymm0
+; SKX-NEXT:    vblendmps (%rdi), %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <8 x i32> %trigger, zeroinitializer
   %res = call <8 x float> @llvm.masked.load.v8f32.p0v8f32(<8 x float>* %addr, i32 32, <8 x i1>%mask, <8 x float>%dst)
@@ -293,16 +288,15 @@
 ; AVX512F-NEXT:    vptestmq %zmm0, %zmm0, %k0
 ; AVX512F-NEXT:    kshiftlw $8, %k0, %k0
 ; AVX512F-NEXT:    kshiftrw $8, %k0, %k1
-; AVX512F-NEXT:    vmovdqu32 (%rdi), %zmm1 {%k1}
-; AVX512F-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512F-NEXT:    vpblendmd (%rdi), %zmm1, %zmm0 {%k1}
+; AVX512F-NEXT:    ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill>
 ; AVX512F-NEXT:    retq
 ;
 ; SKX-LABEL: test11b:
 ; SKX:       ## BB#0:
 ; SKX-NEXT:    vpsllw $15, %xmm0, %xmm0
 ; SKX-NEXT:    vpmovw2m %xmm0, %k1
-; SKX-NEXT:    vmovdqu32 (%rdi), %ymm1 {%k1}
-; SKX-NEXT:    vmovdqa %ymm1, %ymm0
+; SKX-NEXT:    vpblendmd (%rdi), %ymm1, %ymm0 {%k1}
 ; SKX-NEXT:    retq
   %res = call <8 x i32> @llvm.masked.load.v8i32.p0v8i32(<8 x i32>* %addr, i32 4, <8 x i1>%mask, <8 x i32>%dst)
   ret <8 x i32> %res
@@ -557,8 +551,7 @@
 ; SKX-NEXT:    vpcmpeqq %xmm2, %xmm0, %k0
 ; SKX-NEXT:    kshiftlw $14, %k0, %k0
 ; SKX-NEXT:    kshiftrw $14, %k0, %k1
-; SKX-NEXT:    vmovups (%rdi), %xmm1 {%k1}
-; SKX-NEXT:    vmovaps %xmm1, %xmm0
+; SKX-NEXT:    vblendmps (%rdi), %xmm1, %xmm0 {%k1}
 ; SKX-NEXT:    retq
   %mask = icmp eq <2 x i32> %trigger, zeroinitializer
   %res = call <2 x float> @llvm.masked.load.v2f32.p0v2f32(<2 x float>* %addr, i32 4, <2 x i1>%mask, <2 x float>%dst)
Index: llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v16.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v16.ll
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-128-v16.ll
@@ -425,8 +425,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $-21846, %ax # imm = 0xAAAA
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu8 %xmm0, %xmm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %xmm1, %xmm0
+; AVX512VL-NEXT:    vpblendmb %xmm0, %xmm1, %xmm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
   ret <16 x i8> %shuffle
@@ -466,8 +465,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $-30584, %ax # imm = 0x8888
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu8 %xmm0, %xmm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %xmm1, %xmm0
+; AVX512VL-NEXT:    vpblendmb %xmm0, %xmm1, %xmm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 4, i32 5, i32 6, i32 23, i32 8, i32 9, i32 10, i32 27, i32 12, i32 13, i32 14, i32 31>
   ret <16 x i8> %shuffle
@@ -526,8 +524,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $-28528, %ax # imm = 0x9090
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu8 %xmm0, %xmm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %xmm1, %xmm0
+; AVX512VL-NEXT:    vpblendmb %xmm0, %xmm1, %xmm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 5, i32 6, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 13, i32 14, i32 31>
   ret <16 x i8> %shuffle
Index: llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -719,8 +719,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $-32768, %ax # imm = 0x8000
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512VL-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 31>
   ret <16 x i16> %shuffle
@@ -745,8 +744,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $1, %ax
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512VL-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 16, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   ret <16 x i16> %shuffle
@@ -771,8 +769,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $21930, %ax # imm = 0x55AA
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512VL-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 24, i32 9, i32 26, i32 11, i32 28, i32 13, i32 30, i32 15>
   ret <16 x i16> %shuffle
@@ -797,8 +794,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movw $-21931, %ax # imm = 0xAA55
 ; AVX512VL-NEXT:    kmovw %eax, %k1
-; AVX512VL-NEXT:    vmovdqu16 %ymm0, %ymm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512VL-NEXT:    vpblendmw %ymm0, %ymm1, %ymm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 16, i32 1, i32 18, i32 3, i32 20, i32 5, i32 22, i32 7, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
   ret <16 x i16> %shuffle
Index: llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v32.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -1036,8 +1036,7 @@
 ; AVX512VL:       # BB#0:
 ; AVX512VL-NEXT:    movl $-1431655766, %eax # imm = 0xAAAAAAAA
 ; AVX512VL-NEXT:    kmovd %eax, %k1
-; AVX512VL-NEXT:    vmovdqu8 %ymm0, %ymm1 {%k1}
-; AVX512VL-NEXT:    vmovdqa %ymm1, %ymm0
+; AVX512VL-NEXT:    vpblendmb %ymm0, %ymm1, %ymm0 {%k1}
 ; AVX512VL-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> %b, <32 x i32> <i32 0, i32 33, i32 2, i32 35, i32 4, i32 37, i32 6, i32 39, i32 8, i32 41, i32 10, i32 43, i32 12, i32 45, i32 14, i32 47, i32 16, i32 49, i32 18, i32 51, i32 20, i32 53, i32 22, i32 55, i32 24, i32 57, i32 26, i32 59, i32 28, i32 61, i32 30, i32 63>
   ret <32 x i8> %shuffle
Index: llvm/trunk/test/CodeGen/X86/vector-shuffle-512-v16.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-512-v16.ll
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-512-v16.ll
@@ -254,8 +254,7 @@
 ; ALL:       # BB#0:
 ; ALL-NEXT:    movw $8, %ax
 ; ALL-NEXT:    kmovw %eax, %k1
-; ALL-NEXT:    vmovdqa32 %zmm0, %zmm1 {%k1}
-; ALL-NEXT:    vmovdqa64 %zmm1, %zmm0
+; ALL-NEXT:    vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
 ; ALL-NEXT:    retq
   %c = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <16 x i32> %c
Index: llvm/trunk/test/CodeGen/X86/vector-shuffle-masked.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/vector-shuffle-masked.ll
+++ llvm/trunk/test/CodeGen/X86/vector-shuffle-masked.ll
@@ -216,8 +216,7 @@
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[1,2,3,0]
 ; CHECK-NEXT:    kmovb %edi, %k1
-; CHECK-NEXT:    vmovdqa32 %ymm0, %ymm1 {%k1}
-; CHECK-NEXT:    vmovdqa %ymm1, %ymm0
+; CHECK-NEXT:    vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
 ; CHECK-NEXT:    retq
   %shuffle = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1>
   %mask.cast = bitcast i8 %mask to <8 x i1>