Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -17052,8 +17052,8 @@
 
 /// Turns an ISD::CondCode into a value suitable for SSE floating-point mask
 /// CMPs.
-static int translateX86FSETCC(ISD::CondCode SetCCOpcode, SDValue &Op0,
-                              SDValue &Op1) {
+static unsigned translateX86FSETCC(ISD::CondCode SetCCOpcode, SDValue &Op0,
+                                   SDValue &Op1) {
   unsigned SSECC;
   bool Swap = false;
 
@@ -17086,8 +17086,8 @@
   case ISD::SETULT: Swap = true; LLVM_FALLTHROUGH;
   case ISD::SETUGT: SSECC = 6; break;
   case ISD::SETO:   SSECC = 7; break;
-  case ISD::SETUEQ:
-  case ISD::SETONE: SSECC = 8; break;
+  case ISD::SETUEQ: SSECC = 8; break;
+  case ISD::SETONE: SSECC = 12; break;
   }
   if (Swap)
     std::swap(Op0, Op1);
@@ -17269,9 +17269,10 @@
     // emit two comparisons and a logic op to tie them together.
     // TODO: This can be avoided if Intel (and only Intel as of 2016) AVX is
     // available.
+    // If we are using an AVX512 instruction we can use the extended predicate.
     SDValue Cmp;
     unsigned SSECC = translateX86FSETCC(Cond, Op0, Op1);
-    if (SSECC == 8) {
+    if (SSECC >= 8 && Opc == X86ISD::CMPP) {
       // LLVM predicate is SETUEQ or SETONE.
       unsigned CC0, CC1;
       unsigned CombineOpc;
@@ -17709,17 +17710,17 @@
        (Subtarget.hasSSE1() && VT == MVT::f32)) &&
       VT == Cond.getOperand(0).getSimpleValueType() && Cond->hasOneUse()) {
     SDValue CondOp0 = Cond.getOperand(0), CondOp1 = Cond.getOperand(1);
-    int SSECC = translateX86FSETCC(
+    unsigned SSECC = translateX86FSETCC(
         cast<CondCodeSDNode>(Cond.getOperand(2))->get(), CondOp0, CondOp1);
 
-    if (SSECC != 8) {
-      if (Subtarget.hasAVX512()) {
-        SDValue Cmp = DAG.getNode(X86ISD::FSETCCM, DL, MVT::v1i1, CondOp0,
-                                  CondOp1, DAG.getConstant(SSECC, DL, MVT::i8));
-        assert(!VT.isVector() && "Not a scalar type?");
-        return DAG.getNode(X86ISD::SELECTS, DL, VT, Cmp, Op1, Op2);
-      }
+    if (Subtarget.hasAVX512()) {
+      SDValue Cmp = DAG.getNode(X86ISD::FSETCCM, DL, MVT::v1i1, CondOp0,
+                                CondOp1, DAG.getConstant(SSECC, DL, MVT::i8));
+      assert(!VT.isVector() && "Not a scalar type?");
+      return DAG.getNode(X86ISD::SELECTS, DL, VT, Cmp, Op1, Op2);
+    }
 
+    if (SSECC < 8) {
       SDValue Cmp = DAG.getNode(X86ISD::FSETCC, DL, VT, CondOp0, CondOp1,
                                 DAG.getConstant(SSECC, DL, MVT::i8));
 
Index: test/CodeGen/X86/avx512-mov.ll
===================================================================
--- test/CodeGen/X86/avx512-mov.ll
+++ test/CodeGen/X86/avx512-mov.ll
@@ -424,8 +424,7 @@
 ; CHECK-LABEL: test40:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe8,0x57,0xd2]
-; 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:    vcmpneq_oqps %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0x74,0x48,0xc2,0xca,0x0c]
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0x28,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -439,8 +438,7 @@
 ; CHECK-LABEL: test41:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe8,0x57,0xd2]
-; 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:    vcmpneq_oqps %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0x74,0x48,0xc2,0xca,0x0c]
 ; CHECK-NEXT:    vmovups (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0x10,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -454,8 +452,7 @@
 ; CHECK-LABEL: test42:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0x57,0xc9]
-; CHECK-NEXT:    vcmpordps %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x48,0xc2,0xc9,0x07]
-; CHECK-NEXT:    vcmpneqps %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0xc2,0xc9,0x04]
+; CHECK-NEXT:    vcmpneq_oqps %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x48,0xc2,0xc9,0x0c]
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7c,0xc9,0x28,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -469,8 +466,7 @@
 ; CHECK-LABEL: test43:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorps %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf0,0x57,0xc9]
-; CHECK-NEXT:    vcmpordps %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x48,0xc2,0xc9,0x07]
-; CHECK-NEXT:    vcmpneqps %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0xc2,0xc9,0x04]
+; CHECK-NEXT:    vcmpneq_oqps %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x48,0xc2,0xc9,0x0c]
 ; CHECK-NEXT:    vmovups (%rdi), %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7c,0xc9,0x10,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -484,8 +480,7 @@
 ; CHECK-LABEL: test44:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorpd %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe9,0x57,0xd2]
-; 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:    vcmpneq_oqpd %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0xf5,0x48,0xc2,0xca,0x0c]
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0x28,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
@@ -499,8 +494,7 @@
 ; CHECK-LABEL: test45:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorpd %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe9,0x57,0xd2]
-; 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:    vcmpneq_oqpd %zmm2, %zmm1, %k1 ## encoding: [0x62,0xf1,0xf5,0x48,0xc2,0xca,0x0c]
 ; CHECK-NEXT:    vmovupd (%rdi), %zmm0 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0x10,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
@@ -514,8 +508,7 @@
 ; CHECK-LABEL: test46:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorpd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x57,0xc9]
-; CHECK-NEXT:    vcmpordpd %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0xfd,0x48,0xc2,0xc9,0x07]
-; CHECK-NEXT:    vcmpneqpd %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0xc2,0xc9,0x04]
+; CHECK-NEXT:    vcmpneq_oqpd %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0xfd,0x48,0xc2,0xc9,0x0c]
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0x28,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
@@ -529,8 +522,7 @@
 ; CHECK-LABEL: test47:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vxorpd %xmm1, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x57,0xc9]
-; CHECK-NEXT:    vcmpordpd %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0xfd,0x48,0xc2,0xc9,0x07]
-; CHECK-NEXT:    vcmpneqpd %zmm1, %zmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0xc2,0xc9,0x04]
+; CHECK-NEXT:    vcmpneq_oqpd %zmm1, %zmm0, %k1 ## encoding: [0x62,0xf1,0xfd,0x48,0xc2,0xc9,0x0c]
 ; CHECK-NEXT:    vmovupd (%rdi), %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0x10,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
Index: test/CodeGen/X86/avx512-schedule.ll
===================================================================
--- test/CodeGen/X86/avx512-schedule.ll
+++ test/CodeGen/X86/avx512-schedule.ll
@@ -5232,8 +5232,7 @@
 ; CHECK-LABEL: mov_test40:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordps %zmm2, %zmm1, %k1
-; CHECK-NEXT:    vcmpneqps %zmm2, %zmm1, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqps %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0 {%k1} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -5247,8 +5246,7 @@
 ; CHECK-LABEL: mov_test41:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordps %zmm2, %zmm1, %k1
-; CHECK-NEXT:    vcmpneqps %zmm2, %zmm1, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqps %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vmovups (%rdi), %zmm0 {%k1} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -5262,8 +5260,7 @@
 ; CHECK-LABEL: mov_test42:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorps %xmm1, %xmm1, %xmm1 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordps %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vcmpneqps %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqps %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0 {%k1} {z} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -5277,8 +5274,7 @@
 ; CHECK-LABEL: mov_test43:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorps %xmm1, %xmm1, %xmm1 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordps %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vcmpneqps %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqps %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vmovups (%rdi), %zmm0 {%k1} {z} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <16 x float> %mask1, zeroinitializer
@@ -5292,8 +5288,7 @@
 ; CHECK-LABEL: mov_test44:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorpd %xmm2, %xmm2, %xmm2 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordpd %zmm2, %zmm1, %k1
-; CHECK-NEXT:    vcmpneqpd %zmm2, %zmm1, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqpd %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0 {%k1} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
@@ -5307,8 +5302,7 @@
 ; CHECK-LABEL: mov_test45:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorpd %xmm2, %xmm2, %xmm2 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordpd %zmm2, %zmm1, %k1
-; CHECK-NEXT:    vcmpneqpd %zmm2, %zmm1, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqpd %zmm2, %zmm1, %k1
 ; CHECK-NEXT:    vmovupd (%rdi), %zmm0 {%k1} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
@@ -5322,8 +5316,7 @@
 ; CHECK-LABEL: mov_test46:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorpd %xmm1, %xmm1, %xmm1 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordpd %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vcmpneqpd %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqpd %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0 {%k1} {z} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
@@ -5337,8 +5330,7 @@
 ; CHECK-LABEL: mov_test47:
 ; CHECK:       # BB#0:
 ; CHECK-NEXT:    vxorpd %xmm1, %xmm1, %xmm1 # sched: [1:0.50]
-; CHECK-NEXT:    vcmpordpd %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vcmpneqpd %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    vcmpneq_oqpd %zmm1, %zmm0, %k1
 ; CHECK-NEXT:    vmovupd (%rdi), %zmm0 {%k1} {z} # sched: [5:0.50]
 ; CHECK-NEXT:    retq # sched: [2:1.00]
   %mask = fcmp one <8 x double> %mask1, zeroinitializer
Index: test/CodeGen/X86/avx512vl-mov.ll
===================================================================
--- test/CodeGen/X86/avx512vl-mov.ll
+++ test/CodeGen/X86/avx512vl-mov.ll
@@ -277,8 +277,7 @@
 ; CHECK-LABEL: test_256_25:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0xef,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:    vcmpneq_oqps %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf1,0x74,0x28,0xc2,0xca,0x0c]
 ; CHECK-NEXT:    vmovaps (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x29,0x28,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x float> %mask1, zeroinitializer
@@ -292,8 +291,7 @@
 ; CHECK-LABEL: test_256_26:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0xef,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:    vcmpneq_oqps %ymm2, %ymm1, %k1 ## encoding: [0x62,0xf1,0x74,0x28,0xc2,0xca,0x0c]
 ; CHECK-NEXT:    vmovups (%rdi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7c,0x29,0x10,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x float> %mask1, zeroinitializer
@@ -307,8 +305,7 @@
 ; CHECK-LABEL: test_256_27:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0xef,0xc9]
-; CHECK-NEXT:    vcmpordps %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x28,0xc2,0xc9,0x07]
-; CHECK-NEXT:    vcmpneqps %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x29,0xc2,0xc9,0x04]
+; CHECK-NEXT:    vcmpneq_oqps %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x28,0xc2,0xc9,0x0c]
 ; CHECK-NEXT:    vmovaps (%rdi), %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7c,0xa9,0x28,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x float> %mask1, zeroinitializer
@@ -322,8 +319,7 @@
 ; CHECK-LABEL: test_256_28:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpxor %xmm1, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0xef,0xc9]
-; CHECK-NEXT:    vcmpordps %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x28,0xc2,0xc9,0x07]
-; CHECK-NEXT:    vcmpneqps %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x29,0xc2,0xc9,0x04]
+; CHECK-NEXT:    vcmpneq_oqps %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf1,0x7c,0x28,0xc2,0xc9,0x0c]
 ; CHECK-NEXT:    vmovups (%rdi), %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7c,0xa9,0x10,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %mask = fcmp one <8 x float> %mask1, zeroinitializer
Index: test/CodeGen/X86/commute-fcmp.ll
===================================================================
--- test/CodeGen/X86/commute-fcmp.ll
+++ test/CodeGen/X86/commute-fcmp.ll
@@ -121,9 +121,7 @@
 ; AVX512-LABEL: commute_cmpps_ueq:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovaps (%rdi), %xmm1
-; AVX512-NEXT:    vcmpeqps %xmm0, %xmm1, %k0
-; AVX512-NEXT:    vcmpunordps %xmm0, %xmm1, %k1
-; AVX512-NEXT:    korw %k0, %k1, %k1
+; AVX512-NEXT:    vcmpeq_uqps %xmm0, %xmm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
 ; AVX512-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -154,8 +152,7 @@
 ; AVX512-LABEL: commute_cmpps_one:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovaps (%rdi), %xmm1
-; AVX512-NEXT:    vcmpordps %xmm0, %xmm1, %k1
-; AVX512-NEXT:    vcmpneqps %xmm0, %xmm1, %k1 {%k1}
+; AVX512-NEXT:    vcmpneq_oqps %xmm0, %xmm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
 ; AVX512-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -341,9 +338,7 @@
 ; AVX512-LABEL: commute_cmpps_ueq_ymm:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovaps (%rdi), %ymm1
-; AVX512-NEXT:    vcmpeqps %ymm0, %ymm1, %k0
-; AVX512-NEXT:    vcmpunordps %ymm0, %ymm1, %k1
-; AVX512-NEXT:    korw %k0, %k1, %k1
+; AVX512-NEXT:    vcmpeq_uqps %ymm0, %ymm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -379,8 +374,7 @@
 ; AVX512-LABEL: commute_cmpps_one_ymm:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovaps (%rdi), %ymm1
-; AVX512-NEXT:    vcmpordps %ymm0, %ymm1, %k1
-; AVX512-NEXT:    vcmpneqps %ymm0, %ymm1, %k1 {%k1}
+; AVX512-NEXT:    vcmpneq_oqps %ymm0, %ymm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -545,9 +539,7 @@
 ; AVX512-LABEL: commute_cmppd_ueq:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovapd (%rdi), %xmm1
-; AVX512-NEXT:    vcmpeqpd %xmm0, %xmm1, %k0
-; AVX512-NEXT:    vcmpunordpd %xmm0, %xmm1, %k1
-; AVX512-NEXT:    korw %k0, %k1, %k1
+; AVX512-NEXT:    vcmpeq_uqpd %xmm0, %xmm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
 ; AVX512-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -578,8 +570,7 @@
 ; AVX512-LABEL: commute_cmppd_one:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovapd (%rdi), %xmm1
-; AVX512-NEXT:    vcmpordpd %xmm0, %xmm1, %k1
-; AVX512-NEXT:    vcmpneqpd %xmm0, %xmm1, %k1 {%k1}
+; AVX512-NEXT:    vcmpneq_oqpd %xmm0, %xmm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
 ; AVX512-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -788,9 +779,7 @@
 ; AVX512-LABEL: commute_cmppd_ueq_ymmm:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovapd (%rdi), %ymm1
-; AVX512-NEXT:    vcmpeqpd %ymm0, %ymm1, %k0
-; AVX512-NEXT:    vcmpunordpd %ymm0, %ymm1, %k1
-; AVX512-NEXT:    korw %k0, %k1, %k1
+; AVX512-NEXT:    vcmpeq_uqpd %ymm0, %ymm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
 ; AVX512-NEXT:    retq
@@ -826,8 +815,7 @@
 ; AVX512-LABEL: commute_cmppd_one_ymmm:
 ; AVX512:       # BB#0:
 ; AVX512-NEXT:    vmovapd (%rdi), %ymm1
-; AVX512-NEXT:    vcmpordpd %ymm0, %ymm1, %k1
-; AVX512-NEXT:    vcmpneqpd %ymm0, %ymm1, %k1 {%k1}
+; AVX512-NEXT:    vcmpneq_oqpd %ymm0, %ymm1, %k1
 ; AVX512-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
 ; AVX512-NEXT:    retq