[AVX-512] Remove and autoupgrade the masked integer compare intrinsics

[craig.topper]

Summary

These intrinsics aren't used by clang and haven't been for a while.

There's some really terrible codegen in the 32-bit target for avx512bw due to i64 not being legal. But as I said these intrinsics aren't used by clang even before this patch so this codegen reflects our clang behavior today.

[RKSimon]

LGTM

llvm/trunk/include/llvm/IR/IntrinsicsX86.td

 let TargetPrefix = "x86" in {
   // 512-bit
   def int_x86_avx512_vcomi_sd : GCCBuiltin<"__builtin_ia32_vcomisd">,
               Intrinsic<[llvm_i32_ty], [llvm_v2f64_ty,
                          llvm_v2f64_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
   def int_x86_avx512_vcomi_ss : GCCBuiltin<"__builtin_ia32_vcomiss">,
               Intrinsic<[llvm_i32_ty], [llvm_v4f32_ty,
                          llvm_v4f32_ty, llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
-
-  def int_x86_avx512_mask_cmp_b_512:
-        Intrinsic<[llvm_i64_ty], [llvm_v64i8_ty, llvm_v64i8_ty, llvm_i32_ty,
-                  llvm_i64_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_w_512:
-        Intrinsic<[llvm_i32_ty], [llvm_v32i16_ty, llvm_v32i16_ty, llvm_i32_ty,
-                  llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_d_512:
-        Intrinsic<[llvm_i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty, llvm_i32_ty,
-                  llvm_i16_ty], [IntrNoMem ]>;
-  def int_x86_avx512_mask_cmp_q_512:
-        Intrinsic<[llvm_i8_ty], [llvm_v8i64_ty, llvm_v8i64_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-
-  def int_x86_avx512_mask_ucmp_b_512:
-        Intrinsic<[llvm_i64_ty], [llvm_v64i8_ty, llvm_v64i8_ty, llvm_i32_ty,
-                  llvm_i64_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_w_512:
-        Intrinsic<[llvm_i32_ty], [llvm_v32i16_ty, llvm_v32i16_ty, llvm_i32_ty,
-                  llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_d_512:
-        Intrinsic<[llvm_i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty, llvm_i32_ty,
-                  llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_q_512:
-        Intrinsic<[llvm_i8_ty], [llvm_v8i64_ty, llvm_v8i64_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-
-  // 256-bit
-  def int_x86_avx512_mask_cmp_b_256:
-        Intrinsic<[llvm_i32_ty], [llvm_v32i8_ty, llvm_v32i8_ty, llvm_i32_ty,
-                  llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_w_256:
-        Intrinsic<[llvm_i16_ty], [llvm_v16i16_ty, llvm_v16i16_ty, llvm_i32_ty,
-                  llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_d_256:
-        Intrinsic<[llvm_i8_ty], [llvm_v8i32_ty, llvm_v8i32_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_q_256:
-        Intrinsic<[llvm_i8_ty], [llvm_v4i64_ty, llvm_v4i64_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-
-  def int_x86_avx512_mask_ucmp_b_256:
-        Intrinsic<[llvm_i32_ty], [llvm_v32i8_ty, llvm_v32i8_ty, llvm_i32_ty,
-                  llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_w_256:
-        Intrinsic<[llvm_i16_ty], [llvm_v16i16_ty, llvm_v16i16_ty, llvm_i32_ty,
-                  llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_d_256:
-        Intrinsic<[llvm_i8_ty], [llvm_v8i32_ty, llvm_v8i32_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_q_256:
-        Intrinsic<[llvm_i8_ty], [llvm_v4i64_ty, llvm_v4i64_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-
-  // 128-bit
-  def int_x86_avx512_mask_cmp_b_128:
-        Intrinsic<[llvm_i16_ty], [llvm_v16i8_ty, llvm_v16i8_ty, llvm_i32_ty,
-                  llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_w_128:
-        Intrinsic<[llvm_i8_ty], [llvm_v8i16_ty, llvm_v8i16_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_d_128:
-        Intrinsic<[llvm_i8_ty], [llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_cmp_q_128:
-        Intrinsic<[llvm_i8_ty], [llvm_v2i64_ty, llvm_v2i64_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-
-  def int_x86_avx512_mask_ucmp_b_128:
-        Intrinsic<[llvm_i16_ty], [llvm_v16i8_ty, llvm_v16i8_ty, llvm_i32_ty,
-                  llvm_i16_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_w_128:
-        Intrinsic<[llvm_i8_ty], [llvm_v8i16_ty, llvm_v8i16_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_d_128:
-        Intrinsic<[llvm_i8_ty], [llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx512_mask_ucmp_q_128:
-        Intrinsic<[llvm_i8_ty], [llvm_v2i64_ty, llvm_v2i64_ty, llvm_i32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
 }
 
 // Compress, Expand
 let TargetPrefix = "x86" in {
   def int_x86_avx512_mask_compress_ps_512 :
                              GCCBuiltin<"__builtin_ia32_compresssf512_mask">,
         Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_v16f32_ty,
                    llvm_i16_ty], [IntrNoMem]>;

llvm/trunk/lib/IR/AutoUpgrade.cpp

if (Name.startswith("sse2.pcmpeq.") || // Added in 3.1
       Name.startswith("avx512.mask.cvtdq2pd.") || // Added in 4.0
       Name.startswith("avx512.mask.cvtudq2pd.") || // Added in 4.0
       Name.startswith("avx512.mask.pmul.dq.") || // Added in 4.0
       Name.startswith("avx512.mask.pmulu.dq.") || // Added in 4.0
       Name.startswith("avx512.mask.packsswb.") || // Added in 5.0
       Name.startswith("avx512.mask.packssdw.") || // Added in 5.0
       Name.startswith("avx512.mask.packuswb.") || // Added in 5.0
       Name.startswith("avx512.mask.packusdw.") || // Added in 5.0
+      Name.startswith("avx512.mask.cmp.b") || // Added in 5.0
+      Name.startswith("avx512.mask.cmp.d") || // Added in 5.0
+      Name.startswith("avx512.mask.cmp.q") || // Added in 5.0
+      Name.startswith("avx512.mask.cmp.w") || // Added in 5.0
+      Name.startswith("avx512.mask.ucmp.") || // Added in 5.0
       Name == "avx512.mask.add.pd.128" || // Added in 4.0
       Name == "avx512.mask.add.pd.256" || // Added in 4.0
       Name == "avx512.mask.add.ps.128" || // Added in 4.0
       Name == "avx512.mask.add.ps.256" || // Added in 4.0
       Name == "avx512.mask.div.pd.128" || // Added in 4.0
       Name == "avx512.mask.div.pd.256" || // Added in 4.0
       Name == "avx512.mask.div.ps.128" || // Added in 4.0
       Name == "avx512.mask.div.ps.256" || // Added in 4.0
static Value *upgradeIntMinMax(IRBuilder<> &Builder, CallInst &CI,
 
   if (CI.getNumArgOperands() == 4)
     Res = EmitX86Select(Builder, CI.getArgOperand(3), Res, CI.getArgOperand(2));
 
   return Res;
 }
 
 static Value *upgradeMaskedCompare(IRBuilder<> &Builder, CallInst &CI,
-                                   ICmpInst::Predicate Pred) {
+                                   unsigned CC, bool Signed) {
   Value *Op0 = CI.getArgOperand(0);
   unsigned NumElts = Op0->getType()->getVectorNumElements();
-  Value *Cmp = Builder.CreateICmp(Pred, Op0, CI.getArgOperand(1));
 
-  Value *Mask = CI.getArgOperand(2);
+  Value *Cmp;
+  if (CC == 3) {
+    Cmp = Constant::getNullValue(llvm::VectorType::get(Builder.getInt1Ty(), NumElts));
+  } else if (CC == 7) {
+    Cmp = Constant::getAllOnesValue(llvm::VectorType::get(Builder.getInt1Ty(), NumElts));
+  } else {
+    ICmpInst::Predicate Pred;
+    switch (CC) {
+    default: llvm_unreachable("Unknown condition code");
+    case 0: Pred = ICmpInst::ICMP_EQ;  break;
+    case 1: Pred = Signed ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; break;
+    case 2: Pred = Signed ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE; break;
+    case 4: Pred = ICmpInst::ICMP_NE;  break;
+    case 5: Pred = Signed ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE; break;
+    case 6: Pred = Signed ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; break;
+    }
+    Cmp = Builder.CreateICmp(Pred, Op0, CI.getArgOperand(1));
+  }
+
+  Value *Mask = CI.getArgOperand(CI.getNumArgOperands() - 1);
   const auto *C = dyn_cast<Constant>(Mask);
   if (!C || !C->isAllOnesValue())
     Cmp = Builder.CreateAnd(Cmp, getX86MaskVec(Builder, Mask, NumElts));
 
   if (NumElts < 8) {
     uint32_t Indices[8];
     for (unsigned i = 0; i != NumElts; ++i)
       Indices[i] = i;
if (IsX86 && (Name.startswith("sse2.pcmp") ||
       Value *Elt1 = Builder.CreateExtractElement(CI->getArgOperand(1),
                                                  ConstantInt::get(I32Ty, 0));
       Rep = Builder.CreateInsertElement(CI->getArgOperand(0),
                                         Builder.CreateFDiv(Elt0, Elt1),
                                         ConstantInt::get(I32Ty, 0));
     } else if (IsX86 && Name.startswith("avx512.mask.pcmp")) {
       // "avx512.mask.pcmpeq." or "avx512.mask.pcmpgt."
       bool CmpEq = Name[16] == 'e';
-      Rep = upgradeMaskedCompare(Builder, *CI,
-                                 CmpEq ? ICmpInst::ICMP_EQ
-                                       : ICmpInst::ICMP_SGT);
+      Rep = upgradeMaskedCompare(Builder, *CI, CmpEq ? 0 : 6, true);
+    } else if (IsX86 && Name.startswith("avx512.mask.cmp")) {
+      unsigned Imm = cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue();
+      Rep = upgradeMaskedCompare(Builder, *CI, Imm, true);
+    } else if (IsX86 && Name.startswith("avx512.mask.ucmp")) {
+      unsigned Imm = cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue();
+      Rep = upgradeMaskedCompare(Builder, *CI, Imm, false);
     } else if (IsX86 && (Name == "sse41.pmaxsb" ||
                          Name == "sse2.pmaxs.w" ||
                          Name == "sse41.pmaxsd" ||
                          Name.startswith("avx2.pmaxs") ||
                          Name.startswith("avx512.mask.pmaxs"))) {
       Rep = upgradeIntMinMax(Builder, *CI, ICmpInst::ICMP_SGT);
     } else if (IsX86 && (Name == "sse2.pmaxu.b" ||
                          Name == "sse41.pmaxuw" ||

llvm/trunk/lib/Target/X86/X86IntrinsicsInfo.h

static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_broadcasti32x8_512, BRCST_SUBVEC_TO_VEC,
                      X86ISD::SHUF128, 0),
   X86_INTRINSIC_DATA(avx512_mask_broadcasti64x2_256, BRCST_SUBVEC_TO_VEC,
                      X86ISD::SHUF128, 0),
   X86_INTRINSIC_DATA(avx512_mask_broadcasti64x2_512, BRCST_SUBVEC_TO_VEC,
                      X86ISD::SHUF128, 0),
   X86_INTRINSIC_DATA(avx512_mask_broadcasti64x4_512, BRCST_SUBVEC_TO_VEC,
                      X86ISD::SHUF128, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_b_128,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_b_256,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_b_512,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_d_128,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_d_256,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_d_512,  CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_pd_128, CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_pd_256, CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_pd_512, CMP_MASK_CC, X86ISD::CMPM,
                      X86ISD::CMPM_RND),
   X86_INTRINSIC_DATA(avx512_mask_cmp_ps_128, CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_ps_256, CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_ps_512, CMP_MASK_CC, X86ISD::CMPM,
                      X86ISD::CMPM_RND),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_q_128,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_q_256,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_q_512,  CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_cmp_sd,     CMP_MASK_SCALAR_CC,
                      X86ISD::FSETCCM, X86ISD::FSETCCM_RND),
   X86_INTRINSIC_DATA(avx512_mask_cmp_ss,     CMP_MASK_SCALAR_CC,
                      X86ISD::FSETCCM, X86ISD::FSETCCM_RND),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_w_128,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_w_256,  CMP_MASK_CC, X86ISD::CMPM, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cmp_w_512,  CMP_MASK_CC, X86ISD::CMPM, 0),
   X86_INTRINSIC_DATA(avx512_mask_compress_d_128,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::COMPRESS, 0),
   X86_INTRINSIC_DATA(avx512_mask_compress_d_256,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::COMPRESS, 0),
   X86_INTRINSIC_DATA(avx512_mask_compress_d_512,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::COMPRESS, 0),
   X86_INTRINSIC_DATA(avx512_mask_compress_pd_128, COMPRESS_EXPAND_IN_REG,
                      X86ISD::COMPRESS, 0),
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_sub_pd_512, INTR_TYPE_2OP_MASK, ISD::FSUB,
                      X86ISD::FSUB_RND),
   X86_INTRINSIC_DATA(avx512_mask_sub_ps_512, INTR_TYPE_2OP_MASK, ISD::FSUB,
                      X86ISD::FSUB_RND),
   X86_INTRINSIC_DATA(avx512_mask_sub_sd_round, INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::FSUBS_RND, 0),
   X86_INTRINSIC_DATA(avx512_mask_sub_ss_round, INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::FSUBS_RND, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_b_128,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_b_256,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_b_512,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_d_128,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_d_256,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_d_512,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_q_128,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_q_256,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_q_512,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_w_128,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_w_256,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
-  X86_INTRINSIC_DATA(avx512_mask_ucmp_w_512,    CMP_MASK_CC,  X86ISD::CMPMU, 0),
   X86_INTRINSIC_DATA(avx512_mask_vcvtph2ps_128, INTR_TYPE_1OP_MASK_RM,
                      X86ISD::CVTPH2PS, 0),
   X86_INTRINSIC_DATA(avx512_mask_vcvtph2ps_256, INTR_TYPE_1OP_MASK_RM,
                      X86ISD::CVTPH2PS, 0),
   X86_INTRINSIC_DATA(avx512_mask_vcvtph2ps_512, INTR_TYPE_1OP_MASK_RM,
                      X86ISD::CVTPH2PS, 0),
   X86_INTRINSIC_DATA(avx512_mask_vcvtps2ph_128, INTR_TYPE_2OP_MASK,
                      X86ISD::CVTPS2PH, 0),

llvm/trunk/test/CodeGen/X86/avx512-intrinsics-upgrade.ll

 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vmovntdqa (%rdi), %zmm0
 ; CHECK-NEXT:    retq
   %res = call <8 x i64> @llvm.x86.avx512.movntdqa(i8* %a0)
   ret <8 x i64> %res
 }
 
 declare <8 x i64> @llvm.x86.avx512.movntdqa(i8*) nounwind readonly
+
+define <8 x i16> @test_cmp_d_512(<16 x i32> %a0, <16 x i32> %a1) {
+; CHECK-LABEL: test_cmp_d_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0
+; CHECK-NEXT:    vpcmpgtd %zmm0, %zmm1, %k1
+; CHECK-NEXT:    vpcmpled %zmm1, %zmm0, %k2
+; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k3
+; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k4
+; CHECK-NEXT:    vpcmpgtd %zmm1, %zmm0, %k5
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 -1)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 -1)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 -1)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 -1)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 -1)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 -1)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 -1)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 -1)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+define <8 x i16> @test_mask_cmp_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+; CHECK-LABEL: test_mask_cmp_d_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0 {%k1}
+; CHECK-NEXT:    vpcmpgtd %zmm0, %zmm1, %k2 {%k1}
+; CHECK-NEXT:    vpcmpled %zmm1, %zmm0, %k3 {%k1}
+; CHECK-NEXT:    kxorw %k0, %k0, %k4
+; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k5 {%k1}
+; CHECK-NEXT:    vpcmpled %zmm0, %zmm1, %k6 {%k1}
+; CHECK-NEXT:    vpcmpgtd %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    kmovw %k0, %ecx
+; CHECK-NEXT:    vmovd %ecx, %xmm0
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k6, %eax
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 %mask)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 %mask)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 %mask)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 %mask)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 %mask)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 %mask)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 %mask)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 %mask)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+declare i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32>, <16 x i32>, i32, i16) nounwind readnone
+
+define <8 x i16> @test_ucmp_d_512(<16 x i32> %a0, <16 x i32> %a1) {
+; CHECK-LABEL: test_ucmp_d_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0
+; CHECK-NEXT:    vpcmpltud %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vpcmpleud %zmm1, %zmm0, %k2
+; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k3
+; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k4
+; CHECK-NEXT:    vpcmpnleud %zmm1, %zmm0, %k5
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 -1)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 -1)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 -1)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 -1)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 -1)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 -1)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 -1)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 -1)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+define <8 x i16> @test_mask_ucmp_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
+; CHECK-LABEL: test_mask_ucmp_d_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0 {%k1}
+; CHECK-NEXT:    vpcmpltud %zmm1, %zmm0, %k2 {%k1}
+; CHECK-NEXT:    vpcmpleud %zmm1, %zmm0, %k3 {%k1}
+; CHECK-NEXT:    kxorw %k0, %k0, %k4
+; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k5 {%k1}
+; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k6 {%k1}
+; CHECK-NEXT:    vpcmpnleud %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    kmovw %k0, %ecx
+; CHECK-NEXT:    vmovd %ecx, %xmm0
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k6, %eax
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 %mask)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 %mask)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 %mask)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 %mask)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 %mask)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 %mask)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 %mask)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 %mask)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+declare i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32>, <16 x i32>, i32, i16) nounwind readnone
+
+define <8 x i8> @test_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
+; CHECK-LABEL: test_cmp_q_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0
+; CHECK-NEXT:    vpcmpgtq %zmm0, %zmm1, %k1
+; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k2
+; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k3
+; CHECK-NEXT:    vpcmpleq %zmm0, %zmm1, %k4
+; CHECK-NEXT:    vpcmpgtq %zmm1, %zmm0, %k5
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_cmp_q_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
+; CHECK-NEXT:    vpcmpgtq %zmm0, %zmm1, %k2 {%k1}
+; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k3 {%k1}
+; CHECK-NEXT:    kxorw %k0, %k0, %k4
+; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k5 {%k1}
+; CHECK-NEXT:    vpcmpleq %zmm0, %zmm1, %k6 {%k1}
+; CHECK-NEXT:    vpcmpgtq %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    kmovw %k0, %ecx
+; CHECK-NEXT:    vmovd %ecx, %xmm0
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k6, %eax
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64>, <8 x i64>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
+; CHECK-LABEL: test_ucmp_q_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0
+; CHECK-NEXT:    vpcmpltuq %zmm1, %zmm0, %k1
+; CHECK-NEXT:    vpcmpleuq %zmm1, %zmm0, %k2
+; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k3
+; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k4
+; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k5
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0
+; CHECK-NEXT:    kmovw %k0, %eax
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_ucmp_q_512:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
+; CHECK-NEXT:    vpcmpltuq %zmm1, %zmm0, %k2 {%k1}
+; CHECK-NEXT:    vpcmpleuq %zmm1, %zmm0, %k3 {%k1}
+; CHECK-NEXT:    kxorw %k0, %k0, %k4
+; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k5 {%k1}
+; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k6 {%k1}
+; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k2, %eax
+; CHECK-NEXT:    kmovw %k0, %ecx
+; CHECK-NEXT:    vmovd %ecx, %xmm0
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k3, %eax
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k4, %eax
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k5, %eax
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k6, %eax
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    kmovw %k1, %eax
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
+; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
+; CHECK-NEXT:    retq
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64>, <8 x i64>, i32, i8) nounwind readnone

llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll

 ; CHECK-NEXT:    vmovss %xmm0, (%rdi) {%k1}
 ; CHECK-NEXT:    retq
  call void @llvm.x86.avx512.mask.store.ss(i8* %ptr, <4 x float> %data, i8 %mask)
  ret void
 }
 
 declare void @llvm.x86.avx512.mask.store.ss(i8*, <4 x float>, i8 )
 
-define <8 x i16> @test_cmp_d_512(<16 x i32> %a0, <16 x i32> %a1) {
-; CHECK-LABEL: test_cmp_d_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0
-; CHECK-NEXT:    vpcmpltd %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vpcmpled %zmm1, %zmm0, %k2
-; CHECK-NEXT:    vpcmpunordd %zmm1, %zmm0, %k3
-; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k4
-; CHECK-NEXT:    vpcmpnltd %zmm1, %zmm0, %k5
-; CHECK-NEXT:    vpcmpnled %zmm1, %zmm0, %k6
-; CHECK-NEXT:    vpcmpordd %zmm1, %zmm0, %k7
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 -1)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 -1)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 -1)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 -1)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 -1)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 -1)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 -1)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 -1)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-define <8 x i16> @test_mask_cmp_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
-; CHECK-LABEL: test_mask_cmp_d_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1
-; CHECK-NEXT:    vpcmpeqd %zmm1, %zmm0, %k0 {%k1}
-; CHECK-NEXT:    vpcmpltd %zmm1, %zmm0, %k2 {%k1}
-; CHECK-NEXT:    vpcmpled %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    vpcmpunordd %zmm1, %zmm0, %k4 {%k1}
-; CHECK-NEXT:    vpcmpneqd %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpnltd %zmm1, %zmm0, %k6 {%k1}
-; CHECK-NEXT:    vpcmpnled %zmm1, %zmm0, %k7 {%k1}
-; CHECK-NEXT:    vpcmpordd %zmm1, %zmm0, %k1 {%k1}
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 %mask)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 %mask)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 %mask)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 %mask)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 %mask)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 %mask)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 %mask)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 %mask)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-declare i16 @llvm.x86.avx512.mask.cmp.d.512(<16 x i32>, <16 x i32>, i32, i16) nounwind readnone
-
-define <8 x i16> @test_ucmp_d_512(<16 x i32> %a0, <16 x i32> %a1) {
-; CHECK-LABEL: test_ucmp_d_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequd %zmm1, %zmm0, %k0
-; CHECK-NEXT:    vpcmpltud %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vpcmpleud %zmm1, %zmm0, %k2
-; CHECK-NEXT:    vpcmpunordud %zmm1, %zmm0, %k3
-; CHECK-NEXT:    vpcmpnequd %zmm1, %zmm0, %k4
-; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k5
-; CHECK-NEXT:    vpcmpnleud %zmm1, %zmm0, %k6
-; CHECK-NEXT:    vpcmpordud %zmm1, %zmm0, %k7
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 -1)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 -1)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 -1)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 -1)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 -1)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 -1)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 -1)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 -1)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-define <8 x i16> @test_mask_ucmp_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
-; CHECK-LABEL: test_mask_ucmp_d_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1
-; CHECK-NEXT:    vpcmpequd %zmm1, %zmm0, %k0 {%k1}
-; CHECK-NEXT:    vpcmpltud %zmm1, %zmm0, %k2 {%k1}
-; CHECK-NEXT:    vpcmpleud %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    vpcmpunordud %zmm1, %zmm0, %k4 {%k1}
-; CHECK-NEXT:    vpcmpnequd %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpnltud %zmm1, %zmm0, %k6 {%k1}
-; CHECK-NEXT:    vpcmpnleud %zmm1, %zmm0, %k7 {%k1}
-; CHECK-NEXT:    vpcmpordud %zmm1, %zmm0, %k1 {%k1}
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 0, i16 %mask)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 1, i16 %mask)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 2, i16 %mask)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 3, i16 %mask)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 4, i16 %mask)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 5, i16 %mask)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 6, i16 %mask)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> %a0, <16 x i32> %a1, i32 7, i16 %mask)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-declare i16 @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32>, <16 x i32>, i32, i16) nounwind readnone
-
-define <8 x i8> @test_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
-; CHECK-LABEL: test_cmp_q_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0
-; CHECK-NEXT:    vpcmpltq %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k2
-; CHECK-NEXT:    vpcmpunordq %zmm1, %zmm0, %k3
-; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k4
-; CHECK-NEXT:    vpcmpnltq %zmm1, %zmm0, %k5
-; CHECK-NEXT:    vpcmpnleq %zmm1, %zmm0, %k6
-; CHECK-NEXT:    vpcmpordq %zmm1, %zmm0, %k7
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_cmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_cmp_q_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1
-; CHECK-NEXT:    vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
-; CHECK-NEXT:    vpcmpltq %zmm1, %zmm0, %k2 {%k1}
-; CHECK-NEXT:    vpcmpleq %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    vpcmpunordq %zmm1, %zmm0, %k4 {%k1}
-; CHECK-NEXT:    vpcmpneqq %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpnltq %zmm1, %zmm0, %k6 {%k1}
-; CHECK-NEXT:    vpcmpnleq %zmm1, %zmm0, %k7 {%k1}
-; CHECK-NEXT:    vpcmpordq %zmm1, %zmm0, %k1 {%k1}
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.cmp.q.512(<8 x i64>, <8 x i64>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1) {
-; CHECK-LABEL: test_ucmp_q_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequq %zmm1, %zmm0, %k0
-; CHECK-NEXT:    vpcmpltuq %zmm1, %zmm0, %k1
-; CHECK-NEXT:    vpcmpleuq %zmm1, %zmm0, %k2
-; CHECK-NEXT:    vpcmpunorduq %zmm1, %zmm0, %k3
-; CHECK-NEXT:    vpcmpnequq %zmm1, %zmm0, %k4
-; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k5
-; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k6
-; CHECK-NEXT:    vpcmporduq %zmm1, %zmm0, %k7
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_ucmp_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_ucmp_q_512:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1
-; CHECK-NEXT:    vpcmpequq %zmm1, %zmm0, %k0 {%k1}
-; CHECK-NEXT:    vpcmpltuq %zmm1, %zmm0, %k2 {%k1}
-; CHECK-NEXT:    vpcmpleuq %zmm1, %zmm0, %k3 {%k1}
-; CHECK-NEXT:    vpcmpunorduq %zmm1, %zmm0, %k4 {%k1}
-; CHECK-NEXT:    vpcmpnequq %zmm1, %zmm0, %k5 {%k1}
-; CHECK-NEXT:    vpcmpnltuq %zmm1, %zmm0, %k6 {%k1}
-; CHECK-NEXT:    vpcmpnleuq %zmm1, %zmm0, %k7 {%k1}
-; CHECK-NEXT:    vpcmporduq %zmm1, %zmm0, %k1 {%k1}
-; CHECK-NEXT:    kmovw %k2, %eax
-; CHECK-NEXT:    kmovw %k0, %ecx
-; CHECK-NEXT:    vmovd %ecx, %xmm0
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k3, %eax
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k4, %eax
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k5, %eax
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k6, %eax
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k7, %eax
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    kmovw %k1, %eax
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; CHECK-NEXT:    retq
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64> %a0, <8 x i64> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.ucmp.q.512(<8 x i64>, <8 x i64>, i32, i8) nounwind readnone
-
 declare <16 x float> @llvm.x86.avx512.mask.sub.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32)
 declare <16 x float> @llvm.x86.avx512.mask.mul.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32)
 declare <8 x double> @llvm.x86.avx512.mask.mul.pd.512(<8 x double>, <8 x double>, <8 x double>, i8, i32)
 
 define <16 x float> @test_vsubps_rn(<16 x float> %a0, <16 x float> %a1) {
 ; CHECK-LABEL: test_vsubps_rn:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vsubps {rn-sae}, %zmm1, %zmm0, %zmm0

llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll

 ; AVX512F-32-NEXT:    vpackuswb (%eax), %zmm0, %zmm0 {%k1} {z}
 ; AVX512F-32-NEXT:    retl
   %b = load <32 x i16>, <32 x i16>* %ptr_b
   %res = call <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16> %a, <32 x i16> %b, <64 x i8> zeroinitializer, i64 %mask)
   ret <64 x i8> %res
 }
 
 declare <64 x i8> @llvm.x86.avx512.mask.packuswb.512(<32 x i16>, <32 x i16>, <64 x i8>, i64)
+
+define i64 @test_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
+; AVX512BW-LABEL: test_cmp_b_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    vpcmpgtb %zmm0, %zmm1, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpleb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpleb %zmm0, %zmm1, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    kxnorq %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_cmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $60, %esp
+; AVX512F-32-NEXT:  .Lcfi4:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 64
+; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpgtb %zmm0, %zmm1, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleb %zmm0, %zmm1, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    addl (%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kxnorq %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $60, %esp
+; AVX512F-32-NEXT:    retl
+  %res0 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 -1)
+  %res1 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 -1)
+  %ret1 = add i64 %res0, %res1
+  %res2 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 -1)
+  %ret2 = add i64 %ret1, %res2
+  %res3 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 -1)
+  %ret3 = add i64 %ret2, %res3
+  %res4 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 -1)
+  %ret4 = add i64 %ret3, %res4
+  %res5 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 -1)
+  %ret5 = add i64 %ret4, %res5
+  %res6 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 -1)
+  %ret6 = add i64 %ret5, %res6
+  %res7 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 -1)
+  %ret7 = add i64 %ret6, %res7
+  ret i64 %ret7
+}
+
+define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+; AVX512BW-LABEL: test_mask_cmp_b_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rdi, %k1
+; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    vpcmpgtb %zmm0, %zmm1, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpleb %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    kxorq %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    orq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    vpcmpleb %zmm0, %zmm1, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpgtb %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    addq %rdi, %rax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_cmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    pushl %ebx
+; AVX512F-32-NEXT:  .Lcfi5:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 8
+; AVX512F-32-NEXT:    pushl %esi
+; AVX512F-32-NEXT:  .Lcfi6:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 12
+; AVX512F-32-NEXT:    subl $60, %esp
+; AVX512F-32-NEXT:  .Lcfi7:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
+; AVX512F-32-NEXT:  .Lcfi8:
+; AVX512F-32-NEXT:    .cfi_offset %esi, -12
+; AVX512F-32-NEXT:  .Lcfi9:
+; AVX512F-32-NEXT:    .cfi_offset %ebx, -8
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm6
+; AVX512F-32-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $5, %al
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    movb %cl, %bl
+; AVX512F-32-NEXT:    andb $15, %bl
+; AVX512F-32-NEXT:    movb %cl, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k0
+; AVX512F-32-NEXT:    movb %bl, %dl
+; AVX512F-32-NEXT:    shrb $2, %bl
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    movb %cl, %bl
+; AVX512F-32-NEXT:    shrb $4, %bl
+; AVX512F-32-NEXT:    shrb $3, %dl
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    vpsllw $8, %xmm2, %xmm2
+; AVX512F-32-NEXT:    kmovd %ecx, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastw %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    kmovd %edx, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpslld $24, %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    kmovd %ebx, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastd %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpsllq $40, %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $6, %al
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $7, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllq $56, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %ch, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastq %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    andb $2, %al
+; AVX512F-32-NEXT:    shrb %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %ch, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $61440, %eax # imm = 0xF000
+; AVX512F-32-NEXT:    shrl $12, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastd %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $13, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $49152, %eax # imm = 0xC000
+; AVX512F-32-NEXT:    shrl $14, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $32768, %eax # imm = 0x8000
+; AVX512F-32-NEXT:    shrl $15, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $16, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllw $8, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %bl
+; AVX512F-32-NEXT:    andb $15, %bl
+; AVX512F-32-NEXT:    movb %bl, %dl
+; AVX512F-32-NEXT:    shrb $2, %bl
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    shrb $3, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslld $24, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $4, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastd %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $5, %dl
+; AVX512F-32-NEXT:    andb $1, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllq $40, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $6, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill> %EAX<def>
+; AVX512F-32-NEXT:    shrb $7, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllq $56, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $24, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastq %xmm2, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm3, %ymm4, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $28, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastd %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm4, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    movl %ecx, %esi
+; AVX512F-32-NEXT:    shrl $29, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm2[0,1,2,3],zmm0[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %esi, %eax
+; AVX512F-32-NEXT:    shrl $30, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm1
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm1[0,1,2,3],zmm0[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %esi, %eax
+; AVX512F-32-NEXT:    shrl $31, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm1, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    kmovd %ecx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm7
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm4, %ymm1, %ymm7, %ymm1
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    andb $2, %al
+; AVX512F-32-NEXT:    shrb %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllw $8, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslld $24, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $4, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastd %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $5, %al
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $40, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $6, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $7, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %ch, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastq %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    andb $2, %al
+; AVX512F-32-NEXT:    shrb %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %ch, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $61440, %eax # imm = 0xF000
+; AVX512F-32-NEXT:    shrl $12, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastd %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $13, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $49152, %eax # imm = 0xC000
+; AVX512F-32-NEXT:    shrl $14, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $32768, %eax # imm = 0x8000
+; AVX512F-32-NEXT:    shrl $15, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %ebx
+; AVX512F-32-NEXT:    shrl $16, %ebx
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllw $8, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    andb $15, %al
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $2, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    shrb $3, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslld $24, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    shrb $4, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastd %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    shrb $5, %al
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $40, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    shrb $6, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    # kill: %BL<def> %BL<kill> %EBX<kill> %EBX<def>
+; AVX512F-32-NEXT:    shrb $7, %bl
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $24, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastq %xmm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm4
+; AVX512F-32-NEXT:    vpblendvb %ymm3, %ymm4, %ymm1, %ymm1
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $29, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $28, %eax
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastd %xmm3, %xmm3
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm4, %ymm1, %ymm3, %ymm1
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm4, %ymm1, %ymm3, %ymm1
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $30, %eax
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastw %xmm3, %xmm3
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-32-NEXT:    vpblendvb %ymm2, %ymm1, %ymm3, %ymm1
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $31, %eax
+; AVX512F-32-NEXT:    kshiftlq $1, %k0, %k0
+; AVX512F-32-NEXT:    kshiftrq $1, %k0, %k0
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    kshiftlq $63, %k1, %k1
+; AVX512F-32-NEXT:    korq %k1, %k0, %k1
+; AVX512F-32-NEXT:    vpcmpeqb %zmm6, %zmm5, %k0 {%k1}
+; AVX512F-32-NEXT:    vpcmpgtb %zmm5, %zmm6, %k2 {%k1}
+; AVX512F-32-NEXT:    vpcmpleb %zmm6, %zmm5, %k3 {%k1}
+; AVX512F-32-NEXT:    vpcmpneqb %zmm6, %zmm5, %k4 {%k1}
+; AVX512F-32-NEXT:    vpcmpleb %zmm5, %zmm6, %k5 {%k1}
+; AVX512F-32-NEXT:    vpcmpgtb %zmm6, %zmm5, %k1 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k2, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k3, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kxorq %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    orl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    orl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovq %k4, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k5, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k1, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl %esi, %eax
+; AVX512F-32-NEXT:    adcxl %ecx, %edx
+; AVX512F-32-NEXT:    addl $60, %esp
+; AVX512F-32-NEXT:    popl %esi
+; AVX512F-32-NEXT:    popl %ebx
+; AVX512F-32-NEXT:    retl
+  %res0 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 %mask)
+  %res1 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 %mask)
+  %ret1 = add i64 %res0, %res1
+  %res2 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 %mask)
+  %ret2 = add i64 %ret1, %res2
+  %res3 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 %mask)
+  %ret3 = add i64 %ret2, %res3
+  %res4 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 %mask)
+  %ret4 = add i64 %ret3, %res4
+  %res5 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 %mask)
+  %ret5 = add i64 %ret4, %res5
+  %res6 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 %mask)
+  %ret6 = add i64 %ret5, %res6
+  %res7 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 %mask)
+  %ret7 = add i64 %ret6, %res7
+  ret i64 %ret7
+}
+
+declare i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8>, <64 x i8>, i32, i64) nounwind readnone
+
+define i64 @test_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
+; AVX512BW-LABEL: test_ucmp_b_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    vpcmpltub %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpleub %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    kxnorq %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_ucmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    subl $60, %esp
+; AVX512F-32-NEXT:  .Lcfi10:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 64
+; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpltub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpleub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    addl (%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kxnorq %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl $60, %esp
+; AVX512F-32-NEXT:    retl
+  %res0 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 -1)
+  %res1 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 -1)
+  %ret1 = add i64 %res0, %res1
+  %res2 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 -1)
+  %ret2 = add i64 %ret1, %res2
+  %res3 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 -1)
+  %ret3 = add i64 %ret2, %res3
+  %res4 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 -1)
+  %ret4 = add i64 %ret3, %res4
+  %res5 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 -1)
+  %ret5 = add i64 %ret4, %res5
+  %res6 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 -1)
+  %ret6 = add i64 %ret5, %res6
+  %res7 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 -1)
+  %ret7 = add i64 %ret6, %res7
+  ret i64 %ret7
+}
+
+define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
+; AVX512BW-LABEL: test_mask_x86_avx512_ucmp_b_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovq %rdi, %k1
+; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    vpcmpltub %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpleub %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    kxorq %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    orq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rcx
+; AVX512BW-NEXT:    addq %rax, %rcx
+; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovq %k0, %rax
+; AVX512BW-NEXT:    addq %rcx, %rax
+; AVX512BW-NEXT:    addq %rdi, %rax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_x86_avx512_ucmp_b_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    pushl %ebx
+; AVX512F-32-NEXT:  .Lcfi11:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 8
+; AVX512F-32-NEXT:    pushl %esi
+; AVX512F-32-NEXT:  .Lcfi12:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 12
+; AVX512F-32-NEXT:    subl $60, %esp
+; AVX512F-32-NEXT:  .Lcfi13:
+; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
+; AVX512F-32-NEXT:  .Lcfi14:
+; AVX512F-32-NEXT:    .cfi_offset %esi, -12
+; AVX512F-32-NEXT:  .Lcfi15:
+; AVX512F-32-NEXT:    .cfi_offset %ebx, -8
+; AVX512F-32-NEXT:    vmovdqa64 %zmm1, %zmm6
+; AVX512F-32-NEXT:    vmovdqa64 %zmm0, %zmm5
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $5, %al
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    movb %cl, %bl
+; AVX512F-32-NEXT:    andb $15, %bl
+; AVX512F-32-NEXT:    movb %cl, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k0
+; AVX512F-32-NEXT:    movb %bl, %dl
+; AVX512F-32-NEXT:    shrb $2, %bl
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    movb %cl, %bl
+; AVX512F-32-NEXT:    shrb $4, %bl
+; AVX512F-32-NEXT:    shrb $3, %dl
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    vpsllw $8, %xmm2, %xmm2
+; AVX512F-32-NEXT:    kmovd %ecx, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastw %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    kmovd %edx, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpslld $24, %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    kmovd %ebx, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastd %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm2
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpsllq $40, %xmm3, %xmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm2, %ymm3, %ymm3
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7]
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $6, %al
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $7, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllq $56, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %ch, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastq %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    andb $2, %al
+; AVX512F-32-NEXT:    shrb %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %ch, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $61440, %eax # imm = 0xF000
+; AVX512F-32-NEXT:    shrl $12, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastd %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $13, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $49152, %eax # imm = 0xC000
+; AVX512F-32-NEXT:    shrl $14, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $32768, %eax # imm = 0x8000
+; AVX512F-32-NEXT:    shrl $15, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $16, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllw $8, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %bl
+; AVX512F-32-NEXT:    andb $15, %bl
+; AVX512F-32-NEXT:    movb %bl, %dl
+; AVX512F-32-NEXT:    shrb $2, %bl
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    shrb $3, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslld $24, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $4, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastd %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $5, %dl
+; AVX512F-32-NEXT:    andb $1, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllq $40, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $6, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill> %EAX<def>
+; AVX512F-32-NEXT:    shrb $7, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpsllq $56, %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $24, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastq %xmm2, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastw %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm3, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm3, %ymm4, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $28, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpbroadcastd %xmm2, %xmm2
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm0, %ymm4, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm2, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    movl %ecx, %esi
+; AVX512F-32-NEXT:    shrl $29, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm2
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm1, %ymm0, %ymm2, %ymm2
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm2[0,1,2,3],zmm0[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %esi, %eax
+; AVX512F-32-NEXT:    shrl $30, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm2, %ymm0, %ymm1, %ymm1
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm1[0,1,2,3],zmm0[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %esi, %eax
+; AVX512F-32-NEXT:    shrl $31, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm1, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[4,5,6,7]
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    kmovd %ecx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm7
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm4, %ymm1, %ymm7, %ymm1
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    andb $2, %al
+; AVX512F-32-NEXT:    shrb %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllw $8, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslld $24, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $4, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastd %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $5, %al
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $40, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $6, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %cl, %al
+; AVX512F-32-NEXT:    shrb $7, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %ch, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastq %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    andb $2, %al
+; AVX512F-32-NEXT:    shrb %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %ch, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $61440, %eax # imm = 0xF000
+; AVX512F-32-NEXT:    shrl $12, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastd %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $13, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $49152, %eax # imm = 0xC000
+; AVX512F-32-NEXT:    shrl $14, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    andl $32768, %eax # imm = 0x8000
+; AVX512F-32-NEXT:    shrl $15, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0]
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %ebx
+; AVX512F-32-NEXT:    shrl $16, %ebx
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllw $8, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    andb $15, %al
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    shrb $2, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    shrb $3, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslld $24, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    shrb $4, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastd %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    shrb $5, %al
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $40, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %bl, %al
+; AVX512F-32-NEXT:    shrb $6, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    # kill: %BL<def> %BL<kill> %EBX<kill> %EBX<def>
+; AVX512F-32-NEXT:    shrb $7, %bl
+; AVX512F-32-NEXT:    kmovd %ebx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpsllq $56, %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $24, %eax
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastq %xmm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $2, %dl
+; AVX512F-32-NEXT:    shrb %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    movb %al, %dl
+; AVX512F-32-NEXT:    andb $15, %dl
+; AVX512F-32-NEXT:    movb %dl, %al
+; AVX512F-32-NEXT:    shrb $2, %dl
+; AVX512F-32-NEXT:    kmovd %edx, %k0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm1, %ymm4
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm7, %ymm4, %ymm0, %ymm0
+; AVX512F-32-NEXT:    shrb $3, %al
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm1
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm4
+; AVX512F-32-NEXT:    vpblendvb %ymm3, %ymm4, %ymm1, %ymm1
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $29, %eax
+; AVX512F-32-NEXT:    andb $1, %al
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $28, %eax
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastd %xmm3, %xmm3
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm4, %ymm1, %ymm3, %ymm1
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k1
+; AVX512F-32-NEXT:    vpmovm2b %k1, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2]
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-32-NEXT:    vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255]
+; AVX512F-32-NEXT:    vpblendvb %ymm4, %ymm1, %ymm3, %ymm1
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $30, %eax
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512F-32-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-32-NEXT:    kmovd %eax, %k0
+; AVX512F-32-NEXT:    vpmovm2b %k0, %zmm3
+; AVX512F-32-NEXT:    vpbroadcastw %xmm3, %xmm3
+; AVX512F-32-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX512F-32-NEXT:    vpblendvb %ymm2, %ymm1, %ymm3, %ymm1
+; AVX512F-32-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
+; AVX512F-32-NEXT:    movl %ecx, %eax
+; AVX512F-32-NEXT:    shrl $31, %eax
+; AVX512F-32-NEXT:    kshiftlq $1, %k0, %k0
+; AVX512F-32-NEXT:    kshiftrq $1, %k0, %k0
+; AVX512F-32-NEXT:    kmovd %eax, %k1
+; AVX512F-32-NEXT:    kshiftlq $63, %k1, %k1
+; AVX512F-32-NEXT:    korq %k1, %k0, %k1
+; AVX512F-32-NEXT:    vpcmpeqb %zmm6, %zmm5, %k0 {%k1}
+; AVX512F-32-NEXT:    vpcmpltub %zmm6, %zmm5, %k2 {%k1}
+; AVX512F-32-NEXT:    vpcmpleub %zmm6, %zmm5, %k3 {%k1}
+; AVX512F-32-NEXT:    vpcmpneqb %zmm6, %zmm5, %k4 {%k1}
+; AVX512F-32-NEXT:    vpcmpnltub %zmm6, %zmm5, %k5 {%k1}
+; AVX512F-32-NEXT:    vpcmpnleub %zmm6, %zmm5, %k1 {%k1}
+; AVX512F-32-NEXT:    kmovq %k0, (%esp)
+; AVX512F-32-NEXT:    movl (%esp), %eax
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k2, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k3, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kxorq %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    orl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    orl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    kmovq %k4, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k5, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    kmovq %k1, {{[0-9]+}}(%esp)
+; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
+; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
+; AVX512F-32-NEXT:    addl %esi, %eax
+; AVX512F-32-NEXT:    adcxl %ecx, %edx
+; AVX512F-32-NEXT:    addl $60, %esp
+; AVX512F-32-NEXT:    popl %esi
+; AVX512F-32-NEXT:    popl %ebx
+; AVX512F-32-NEXT:    retl
+  %res0 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 %mask)
+  %res1 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 %mask)
+  %ret1 = add i64 %res0, %res1
+  %res2 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 %mask)
+  %ret2 = add i64 %ret1, %res2
+  %res3 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 %mask)
+  %ret3 = add i64 %ret2, %res3
+  %res4 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 %mask)
+  %ret4 = add i64 %ret3, %res4
+  %res5 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 %mask)
+  %ret5 = add i64 %ret4, %res5
+  %res6 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 %mask)
+  %ret6 = add i64 %ret5, %res6
+  %res7 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 %mask)
+  %ret7 = add i64 %ret6, %res7
+  ret i64 %ret7
+}
+
+declare i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8>, <64 x i8>, i32, i64) nounwind readnone
+
+define i32 @test_cmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
+; AVX512BW-LABEL: test_cmp_w_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    vpcmpgtw %zmm0, %zmm1, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmplew %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmplew %zmm0, %zmm1, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    kxnord %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_cmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpgtw %zmm0, %zmm1, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmplew %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmplew %zmm0, %zmm1, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    kxnord %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    retl
+  %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
+  %res1 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 -1)
+  %ret1 = add i32 %res0, %res1
+  %res2 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 -1)
+  %ret2 = add i32 %ret1, %res2
+  %res3 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 -1)
+  %ret3 = add i32 %ret2, %res3
+  %res4 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 -1)
+  %ret4 = add i32 %ret3, %res4
+  %res5 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 -1)
+  %ret5 = add i32 %ret4, %res5
+  %res6 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 -1)
+  %ret6 = add i32 %ret5, %res6
+  %res7 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 -1)
+  %ret7 = add i32 %ret6, %res7
+  ret i32 %ret7
+}
+
+define i32 @test_mask_cmp_w_512(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+; AVX512BW-LABEL: test_mask_cmp_w_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %edi, %k1
+; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    vpcmpgtw %zmm0, %zmm1, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmplew %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    kxord %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    orl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    vpcmplew %zmm0, %zmm1, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    addl %edi, %eax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_cmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    kmovd %ecx, %k1
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpgtw %zmm0, %zmm1, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %eax, %edx
+; AVX512F-32-NEXT:    vpcmplew %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    kxord %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    orl %eax, %edx
+; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    vpcmplew %zmm0, %zmm1, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %eax, %edx
+; AVX512F-32-NEXT:    vpcmpgtw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    retl
+  %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 %mask)
+  %res1 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 %mask)
+  %ret1 = add i32 %res0, %res1
+  %res2 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 %mask)
+  %ret2 = add i32 %ret1, %res2
+  %res3 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 %mask)
+  %ret3 = add i32 %ret2, %res3
+  %res4 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 %mask)
+  %ret4 = add i32 %ret3, %res4
+  %res5 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 %mask)
+  %ret5 = add i32 %ret4, %res5
+  %res6 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 %mask)
+  %ret6 = add i32 %ret5, %res6
+  %res7 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 %mask)
+  %ret7 = add i32 %ret6, %res7
+  ret i32 %ret7
+}
+
+declare i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16>, <32 x i16>, i32, i32) nounwind readnone
+
+define i32 @test_ucmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
+; AVX512BW-LABEL: test_ucmp_w_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    kxnord %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_ucmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %ecx
+; AVX512F-32-NEXT:    addl %eax, %ecx
+; AVX512F-32-NEXT:    kxnord %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    retl
+  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
+  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 -1)
+  %ret1 = add i32 %res0, %res1
+  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 -1)
+  %ret2 = add i32 %ret1, %res2
+  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 -1)
+  %ret3 = add i32 %ret2, %res3
+  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 -1)
+  %ret4 = add i32 %ret3, %res4
+  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 -1)
+  %ret5 = add i32 %ret4, %res5
+  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 -1)
+  %ret6 = add i32 %ret5, %res6
+  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 -1)
+  %ret7 = add i32 %ret6, %res7
+  ret i32 %ret7
+}
+
+define i32 @test_mask_ucmp_w_512(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
+; AVX512BW-LABEL: test_mask_ucmp_w_512:
+; AVX512BW:       ## BB#0:
+; AVX512BW-NEXT:    kmovd %edi, %k1
+; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    kxord %k0, %k0, %k0
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    orl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512BW-NEXT:    kmovd %k0, %eax
+; AVX512BW-NEXT:    addl %ecx, %eax
+; AVX512BW-NEXT:    addl %edi, %eax
+; AVX512BW-NEXT:    retq
+;
+; AVX512F-32-LABEL: test_mask_ucmp_w_512:
+; AVX512F-32:       # BB#0:
+; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; AVX512F-32-NEXT:    kmovd %ecx, %k1
+; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %eax, %edx
+; AVX512F-32-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    kxord %k0, %k0, %k0
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    orl %eax, %edx
+; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %edx
+; AVX512F-32-NEXT:    addl %eax, %edx
+; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0 {%k1}
+; AVX512F-32-NEXT:    kmovd %k0, %eax
+; AVX512F-32-NEXT:    addl %edx, %eax
+; AVX512F-32-NEXT:    addl %ecx, %eax
+; AVX512F-32-NEXT:    retl
+  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 %mask)
+  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 %mask)
+  %ret1 = add i32 %res0, %res1
+  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 %mask)
+  %ret2 = add i32 %ret1, %res2
+  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 %mask)
+  %ret3 = add i32 %ret2, %res3
+  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 %mask)
+  %ret4 = add i32 %ret3, %res4
+  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 %mask)
+  %ret5 = add i32 %ret4, %res5
+  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 %mask)
+  %ret6 = add i32 %ret5, %res6
+  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 %mask)
+  %ret7 = add i32 %ret6, %res7
+  ret i32 %ret7
+}
+
+declare i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16>, <32 x i16>, i32, i32) nounwind readnone

llvm/trunk/test/CodeGen/X86/avx512bw-intrinsics.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX512BW
 ; RUN: llc < %s -mtriple=i386-unknown-linux-gnu -mcpu=knl -mattr=+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=AVX512F-32
 
-define i64 @test_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
-; AVX512BW-LABEL: test_cmp_b_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    vpcmpltb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpleb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpunordb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpnltb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rdx
-; AVX512BW-NEXT:    addq %rcx, %rdx
-; AVX512BW-NEXT:    vpcmpordb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rdx, %rax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_cmp_b_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    subl $68, %esp
-; AVX512F-32-NEXT:  .Lcfi0:
-; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
-; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpltb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpleb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpunordb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnltb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpordb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, (%esp)
-; AVX512F-32-NEXT:    addl (%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    addl $68, %esp
-; AVX512F-32-NEXT:    retl
-  %res0 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 -1)
-  %res1 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 -1)
-  %ret1 = add i64 %res0, %res1
-  %res2 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 -1)
-  %ret2 = add i64 %ret1, %res2
-  %res3 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 -1)
-  %ret3 = add i64 %ret2, %res3
-  %res4 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 -1)
-  %ret4 = add i64 %ret3, %res4
-  %res5 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 -1)
-  %ret5 = add i64 %ret4, %res5
-  %res6 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 -1)
-  %ret6 = add i64 %ret5, %res6
-  %res7 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 -1)
-  %ret7 = add i64 %ret6, %res7
-  ret i64 %ret7
-}
-
-define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
-; AVX512BW-LABEL: test_mask_cmp_b_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    kmovq %rdi, %k1
-; AVX512BW-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    vpcmpltb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpleb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpunordb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpnltb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rdx
-; AVX512BW-NEXT:    addq %rcx, %rdx
-; AVX512BW-NEXT:    vpcmpordb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rdx, %rax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_mask_cmp_b_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    subl $68, %esp
-; AVX512F-32-NEXT:  .Lcfi1:
-; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
-; AVX512F-32-NEXT:    vpcmpeqb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, (%esp)
-; AVX512F-32-NEXT:    movl (%esp), %eax
-; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpltb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpleb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpunordb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpneqb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnltb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnleb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpordb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    addl $68, %esp
-; AVX512F-32-NEXT:    retl
-  %res0 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 %mask)
-  %res1 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 %mask)
-  %ret1 = add i64 %res0, %res1
-  %res2 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 %mask)
-  %ret2 = add i64 %ret1, %res2
-  %res3 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 %mask)
-  %ret3 = add i64 %ret2, %res3
-  %res4 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 %mask)
-  %ret4 = add i64 %ret3, %res4
-  %res5 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 %mask)
-  %ret5 = add i64 %ret4, %res5
-  %res6 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 %mask)
-  %ret6 = add i64 %ret5, %res6
-  %res7 = call i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 %mask)
-  %ret7 = add i64 %ret6, %res7
-  ret i64 %ret7
-}
-
-declare i64 @llvm.x86.avx512.mask.cmp.b.512(<64 x i8>, <64 x i8>, i32, i64) nounwind readnone
-
-define i64 @test_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1) {
-; AVX512BW-LABEL: test_ucmp_b_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    vpcmpequb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    vpcmpltub %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpleub %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpunordub %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpnequb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rdx
-; AVX512BW-NEXT:    addq %rcx, %rdx
-; AVX512BW-NEXT:    vpcmpordub %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rdx, %rax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_ucmp_b_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    subl $68, %esp
-; AVX512F-32-NEXT:  .Lcfi2:
-; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
-; AVX512F-32-NEXT:    vpcmpequb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpltub %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpleub %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpunordub %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnequb %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpordub %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovq %k0, (%esp)
-; AVX512F-32-NEXT:    addl (%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    addl $68, %esp
-; AVX512F-32-NEXT:    retl
-  %res0 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 -1)
-  %res1 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 -1)
-  %ret1 = add i64 %res0, %res1
-  %res2 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 -1)
-  %ret2 = add i64 %ret1, %res2
-  %res3 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 -1)
-  %ret3 = add i64 %ret2, %res3
-  %res4 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 -1)
-  %ret4 = add i64 %ret3, %res4
-  %res5 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 -1)
-  %ret5 = add i64 %ret4, %res5
-  %res6 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 -1)
-  %ret6 = add i64 %ret5, %res6
-  %res7 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 -1)
-  %ret7 = add i64 %ret6, %res7
-  ret i64 %ret7
-}
-
-define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
-; AVX512BW-LABEL: test_mask_x86_avx512_ucmp_b_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    kmovq %rdi, %k1
-; AVX512BW-NEXT:    vpcmpequb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    vpcmpltub %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpleub %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpunordub %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpnequb %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rcx, %rax
-; AVX512BW-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rcx
-; AVX512BW-NEXT:    addq %rax, %rcx
-; AVX512BW-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rdx
-; AVX512BW-NEXT:    addq %rcx, %rdx
-; AVX512BW-NEXT:    vpcmpordub %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovq %k0, %rax
-; AVX512BW-NEXT:    addq %rdx, %rax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_mask_x86_avx512_ucmp_b_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    subl $68, %esp
-; AVX512F-32-NEXT:  .Lcfi3:
-; AVX512F-32-NEXT:    .cfi_def_cfa_offset 72
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
-; AVX512F-32-NEXT:    vpcmpequb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, (%esp)
-; AVX512F-32-NEXT:    movl (%esp), %eax
-; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpltub %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpleub %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpunordub %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnequb %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnltub %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpnleub %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    vpcmpordub %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT:    addl {{[0-9]+}}(%esp), %eax
-; AVX512F-32-NEXT:    adcxl {{[0-9]+}}(%esp), %edx
-; AVX512F-32-NEXT:    addl $68, %esp
-; AVX512F-32-NEXT:    retl
-  %res0 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 0, i64 %mask)
-  %res1 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 1, i64 %mask)
-  %ret1 = add i64 %res0, %res1
-  %res2 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 2, i64 %mask)
-  %ret2 = add i64 %ret1, %res2
-  %res3 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 3, i64 %mask)
-  %ret3 = add i64 %ret2, %res3
-  %res4 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 4, i64 %mask)
-  %ret4 = add i64 %ret3, %res4
-  %res5 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 5, i64 %mask)
-  %ret5 = add i64 %ret4, %res5
-  %res6 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 6, i64 %mask)
-  %ret6 = add i64 %ret5, %res6
-  %res7 = call i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8> %a0, <64 x i8> %a1, i32 7, i64 %mask)
-  %ret7 = add i64 %ret6, %res7
-  ret i64 %ret7
-}
-
-declare i64 @llvm.x86.avx512.mask.ucmp.b.512(<64 x i8>, <64 x i8>, i32, i64) nounwind readnone
-
-define i32 @test_cmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
-; AVX512BW-LABEL: test_cmp_w_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    vpcmpltw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmplew %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpunordw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpnltw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %edx
-; AVX512BW-NEXT:    addl %ecx, %edx
-; AVX512BW-NEXT:    vpcmpordw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %edx, %eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_cmp_w_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    vpcmpltw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmplew %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpunordw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpnltw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %edx
-; AVX512F-32-NEXT:    addl %ecx, %edx
-; AVX512F-32-NEXT:    vpcmpordw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %edx, %eax
-; AVX512F-32-NEXT:    retl
-  %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
-  %res1 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 -1)
-  %ret1 = add i32 %res0, %res1
-  %res2 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 -1)
-  %ret2 = add i32 %ret1, %res2
-  %res3 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 -1)
-  %ret3 = add i32 %ret2, %res3
-  %res4 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 -1)
-  %ret4 = add i32 %ret3, %res4
-  %res5 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 -1)
-  %ret5 = add i32 %ret4, %res5
-  %res6 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 -1)
-  %ret6 = add i32 %ret5, %res6
-  %res7 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 -1)
-  %ret7 = add i32 %ret6, %res7
-  ret i32 %ret7
-}
-
-define i32 @test_mask_cmp_w_512(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
-; AVX512BW-LABEL: test_mask_cmp_w_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    kmovd %edi, %k1
-; AVX512BW-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    vpcmpltw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmplew %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpunordw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpnltw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %edx
-; AVX512BW-NEXT:    addl %ecx, %edx
-; AVX512BW-NEXT:    vpcmpordw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %edx, %eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_mask_cmp_w_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    vpcmpeqw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    vpcmpltw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmplew %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpunordw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpneqw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpnltw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpnlew %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %edx
-; AVX512F-32-NEXT:    addl %ecx, %edx
-; AVX512F-32-NEXT:    vpcmpordw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %edx, %eax
-; AVX512F-32-NEXT:    retl
-  %res0 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 %mask)
-  %res1 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 %mask)
-  %ret1 = add i32 %res0, %res1
-  %res2 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 %mask)
-  %ret2 = add i32 %ret1, %res2
-  %res3 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 %mask)
-  %ret3 = add i32 %ret2, %res3
-  %res4 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 %mask)
-  %ret4 = add i32 %ret3, %res4
-  %res5 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 %mask)
-  %ret5 = add i32 %ret4, %res5
-  %res6 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 %mask)
-  %ret6 = add i32 %ret5, %res6
-  %res7 = call i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 %mask)
-  %ret7 = add i32 %ret6, %res7
-  ret i32 %ret7
-}
-
-declare i32 @llvm.x86.avx512.mask.cmp.w.512(<32 x i16>, <32 x i16>, i32, i32) nounwind readnone
-
-define i32 @test_ucmp_w_512(<32 x i16> %a0, <32 x i16> %a1) {
-; AVX512BW-LABEL: test_ucmp_w_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    vpcmpequw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpunorduw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpnequw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %edx
-; AVX512BW-NEXT:    addl %ecx, %edx
-; AVX512BW-NEXT:    vpcmporduw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %edx, %eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_ucmp_w_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    vpcmpequw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpunorduw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpnequw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %edx
-; AVX512F-32-NEXT:    addl %ecx, %edx
-; AVX512F-32-NEXT:    vpcmporduw %zmm1, %zmm0, %k0
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %edx, %eax
-; AVX512F-32-NEXT:    retl
-  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 -1)
-  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 -1)
-  %ret1 = add i32 %res0, %res1
-  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 -1)
-  %ret2 = add i32 %ret1, %res2
-  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 -1)
-  %ret3 = add i32 %ret2, %res3
-  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 -1)
-  %ret4 = add i32 %ret3, %res4
-  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 -1)
-  %ret5 = add i32 %ret4, %res5
-  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 -1)
-  %ret6 = add i32 %ret5, %res6
-  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 -1)
-  %ret7 = add i32 %ret6, %res7
-  ret i32 %ret7
-}
-
-define i32 @test_mask_ucmp_w_512(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
-; AVX512BW-LABEL: test_mask_ucmp_w_512:
-; AVX512BW:       ## BB#0:
-; AVX512BW-NEXT:    kmovd %edi, %k1
-; AVX512BW-NEXT:    vpcmpequw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpunorduw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpnequw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %ecx, %eax
-; AVX512BW-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %ecx
-; AVX512BW-NEXT:    addl %eax, %ecx
-; AVX512BW-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %edx
-; AVX512BW-NEXT:    addl %ecx, %edx
-; AVX512BW-NEXT:    vpcmporduw %zmm1, %zmm0, %k0 {%k1}
-; AVX512BW-NEXT:    kmovd %k0, %eax
-; AVX512BW-NEXT:    addl %edx, %eax
-; AVX512BW-NEXT:    retq
-;
-; AVX512F-32-LABEL: test_mask_ucmp_w_512:
-; AVX512F-32:       # BB#0:
-; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
-; AVX512F-32-NEXT:    vpcmpequw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    vpcmpltuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpleuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpunorduw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpnequw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %ecx, %eax
-; AVX512F-32-NEXT:    vpcmpnltuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %ecx
-; AVX512F-32-NEXT:    addl %eax, %ecx
-; AVX512F-32-NEXT:    vpcmpnleuw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %edx
-; AVX512F-32-NEXT:    addl %ecx, %edx
-; AVX512F-32-NEXT:    vpcmporduw %zmm1, %zmm0, %k0 {%k1}
-; AVX512F-32-NEXT:    kmovd %k0, %eax
-; AVX512F-32-NEXT:    addl %edx, %eax
-; AVX512F-32-NEXT:    retl
-  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 0, i32 %mask)
-  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 1, i32 %mask)
-  %ret1 = add i32 %res0, %res1
-  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 2, i32 %mask)
-  %ret2 = add i32 %ret1, %res2
-  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 3, i32 %mask)
-  %ret3 = add i32 %ret2, %res3
-  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 4, i32 %mask)
-  %ret4 = add i32 %ret3, %res4
-  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 5, i32 %mask)
-  %ret5 = add i32 %ret4, %res5
-  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 6, i32 %mask)
-  %ret6 = add i32 %ret5, %res6
-  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16> %a0, <32 x i16> %a1, i32 7, i32 %mask)
-  %ret7 = add i32 %ret6, %res7
-  ret i32 %ret7
-}
-
-declare i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16>, <32 x i16>, i32, i32) nounwind readnone
-
 define <32 x i16> @test_mask_packs_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) {
 ; AVX512BW-LABEL: test_mask_packs_epi32_rr_512:
 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpackssdw %zmm1, %zmm0, %zmm0
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512F-32-LABEL: test_mask_packs_epi32_rr_512:
 ; AVX512F-32:       # BB#0:

 ; AVX512BW-NEXT:    kmovq %rsi, %k1
 ; AVX512BW-NEXT:    kunpckdq %k1, %k0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512F-32-LABEL: test_int_x86_avx512_kunpck_qd:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    subl $12, %esp
-; AVX512F-32-NEXT:  .Lcfi4:
+; AVX512F-32-NEXT:  .Lcfi0:
 ; AVX512F-32-NEXT:    .cfi_def_cfa_offset 16
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k0
 ; AVX512F-32-NEXT:    kmovq %k0, (%esp)
 ; AVX512F-32-NEXT:    movl (%esp), %eax
 ; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; AVX512F-32-NEXT:    addl $12, %esp

 ; AVX512BW:       ## BB#0:
 ; AVX512BW-NEXT:    vpmovb2m %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512F-32-LABEL: test_int_x86_avx512_cvtb2mask_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    subl $12, %esp
-; AVX512F-32-NEXT:  .Lcfi5:
+; AVX512F-32-NEXT:  .Lcfi1:
 ; AVX512F-32-NEXT:    .cfi_def_cfa_offset 16
 ; AVX512F-32-NEXT:    vpmovb2m %zmm0, %k0
 ; AVX512F-32-NEXT:    kmovq %k0, (%esp)
 ; AVX512F-32-NEXT:    movl (%esp), %eax
 ; AVX512F-32-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; AVX512F-32-NEXT:    addl $12, %esp
 ; AVX512F-32-NEXT:    retl
     %res = call i64 @llvm.x86.avx512.cvtb2mask.512(<64 x i8> %x0)

 ; AVX512BW-NEXT:    vptestmb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    addq %rcx, %rax
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512F-32-LABEL: test_int_x86_avx512_ptestm_b_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    subl $20, %esp
-; AVX512F-32-NEXT:  .Lcfi6:
+; AVX512F-32-NEXT:  .Lcfi2:
 ; AVX512F-32-NEXT:    .cfi_def_cfa_offset 24
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
 ; AVX512F-32-NEXT:    vptestmb %zmm1, %zmm0, %k0 {%k1}
 ; AVX512F-32-NEXT:    kmovq %k0, (%esp)
 ; AVX512F-32-NEXT:    vptestmb %zmm1, %zmm0, %k0
 ; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)

 ; AVX512BW-NEXT:    vptestnmb %zmm1, %zmm0, %k0
 ; AVX512BW-NEXT:    kmovq %k0, %rax
 ; AVX512BW-NEXT:    addq %rcx, %rax
 ; AVX512BW-NEXT:    retq
 ;
 ; AVX512F-32-LABEL: test_int_x86_avx512_ptestnm_b_512:
 ; AVX512F-32:       # BB#0:
 ; AVX512F-32-NEXT:    subl $20, %esp
-; AVX512F-32-NEXT:  .Lcfi7:
+; AVX512F-32-NEXT:  .Lcfi3:
 ; AVX512F-32-NEXT:    .cfi_def_cfa_offset 24
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k0
 ; AVX512F-32-NEXT:    kmovd {{[0-9]+}}(%esp), %k1
 ; AVX512F-32-NEXT:    kunpckdq %k0, %k1, %k1
 ; AVX512F-32-NEXT:    vptestnmb %zmm1, %zmm0, %k0 {%k1}
 ; AVX512F-32-NEXT:    kmovq %k0, (%esp)
 ; AVX512F-32-NEXT:    vptestnmb %zmm1, %zmm0, %k0
 ; AVX512F-32-NEXT:    kmovq %k0, {{[0-9]+}}(%esp)

llvm/trunk/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll

 ; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
   ret i16 %res
 }
 
 declare i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16>, <16 x i16>, i16)
 
+define i16 @test_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b) {
+; CHECK-LABEL: test_pcmpeq_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
+  ret i16 %res
+}
+
+define i16 @test_mask_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
+; CHECK-LABEL: test_mask_pcmpeq_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
+  ret i16 %res
+}
+
+declare i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8>, <16 x i8>, i16)
+
+define i8 @test_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b) {
+; CHECK-LABEL: test_pcmpeq_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
+  ret i8 %res
+}
+
+define i8 @test_mask_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
+; CHECK-LABEL: test_mask_pcmpeq_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
+  ret i8 %res
+}
+
+declare i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16>, <8 x i16>, i8)
+
+define i16 @test_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b) {
+; CHECK-LABEL: test_pcmpgt_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
+  ret i16 %res
+}
+
+define i16 @test_mask_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
+; CHECK-LABEL: test_mask_pcmpgt_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
+  ret i16 %res
+}
+
+declare i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8>, <16 x i8>, i16)
+
+define i8 @test_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b) {
+; CHECK-LABEL: test_pcmpgt_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
+  ret i8 %res
+}
+
+define i8 @test_mask_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
+; CHECK-LABEL: test_mask_pcmpgt_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
+  ret i8 %res
+}
+
+declare i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16>, <8 x i16>, i8)
+
 declare <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
 
 define <16 x i8>@test_int_x86_avx512_mask_punpckhb_w_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpunpckhbw %xmm1, %xmm0, %xmm3 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
 ; CHECK-NEXT:    ## xmm3 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]

 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %b = load <16 x i16>, <16 x i16>* %ptr_b
   %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
   ret <32 x i8> %res
 }
 
 declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
 
+define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: test_cmp_b_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
+; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
+; CHECK-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
+; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf3,0x75,0x28,0x3f,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
+; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
+; CHECK-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
+; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
+; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
+; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
+; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
+; CHECK-NEXT:    ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-NEXT:    vmovd %edx, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
+; CHECK-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
+; CHECK-NEXT:    ## xmm1 = xmm1[0],xmm2[0]
+; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
+  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
+  %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
+  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
+  %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
+  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
+  %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
+  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
+  %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
+  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
+  %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
+  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
+  %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
+  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
+  %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
+  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
+  ret <8 x i32> %vec7
+}
+
+define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
+; CHECK-LABEL: test_mask_cmp_b_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
+; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
+; CHECK-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
+; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
+; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
+; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
+; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpcmpleb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x3f,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
+; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
+; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01]
+; CHECK-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
+; CHECK-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
+; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
+; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
+; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
+; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
+; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
+  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
+  %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
+  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
+  %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
+  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
+  %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
+  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
+  %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
+  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
+  %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
+  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
+  %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
+  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
+  %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
+  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
+  ret <8 x i32> %vec7
+}
+
+declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
+
+define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: test_ucmp_b_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
+; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
+; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
+; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
+; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
+; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
+; CHECK-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
+; CHECK-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
+; CHECK-NEXT:    kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
+; CHECK-NEXT:    vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
+; CHECK-NEXT:    vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
+; CHECK-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
+; CHECK-NEXT:    ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; CHECK-NEXT:    vmovd %edx, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
+; CHECK-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
+; CHECK-NEXT:    ## xmm1 = xmm1[0],xmm2[0]
+; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
+  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
+  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
+  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
+  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
+  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
+  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
+  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
+  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
+  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
+  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
+  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
+  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
+  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
+  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
+  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
+  ret <8 x i32> %vec7
+}
+
+define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
+; CHECK-LABEL: test_mask_ucmp_b_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
+; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
+; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
+; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
+; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
+; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
+; CHECK-NEXT:    kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0]
+; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
+; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
+; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
+; CHECK-NEXT:    vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0]
+; CHECK-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01]
+; CHECK-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
+; CHECK-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
+; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
+; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
+; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
+; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
+; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
+  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
+  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
+  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
+  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
+  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
+  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
+  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
+  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
+  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
+  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
+  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
+  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
+  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
+  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
+  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
+  ret <8 x i32> %vec7
+}
+
+declare i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
+
+define <8 x i16> @test_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: test_cmp_w_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
+; CHECK-NEXT:    vpcmpgtw %ymm0, %ymm1, %k1 ## encoding: [0x62,0xf1,0x75,0x28,0x65,0xc8]
+; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
+; CHECK-NEXT:    vpcmplew %ymm0, %ymm1, %k4 ## encoding: [0x62,0xf3,0xf5,0x28,0x3f,0xe0,0x02]
+; CHECK-NEXT:    vpcmpgtw %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf1,0x7d,0x28,0x65,0xe9]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+define <8 x i16> @test_mask_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
+; CHECK-LABEL: test_mask_cmp_w_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
+; CHECK-NEXT:    vpcmpgtw %ymm0, %ymm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
+; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x04]
+; CHECK-NEXT:    vpcmplew %ymm0, %ymm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x3f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpgtw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+declare i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
+
+define <8 x i16> @test_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: test_ucmp_w_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
+; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
+; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe1,0x05]
+; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe9,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+define <8 x i16> @test_mask_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
+; CHECK-LABEL: test_mask_ucmp_w_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
+; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
+; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x04]
+; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+declare i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
+
+define <8 x i16> @test_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
+; CHECK-LABEL: test_cmp_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
+; CHECK-NEXT:    vpcmpgtb %xmm0, %xmm1, %k1 ## encoding: [0x62,0xf1,0x75,0x08,0x64,0xc8]
+; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
+; CHECK-NEXT:    vpcmpleb %xmm0, %xmm1, %k4 ## encoding: [0x62,0xf3,0x75,0x08,0x3f,0xe0,0x02]
+; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf1,0x7d,0x08,0x64,0xe9]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+define <8 x i16> @test_mask_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
+; CHECK-LABEL: test_mask_cmp_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
+; CHECK-NEXT:    vpcmpgtb %xmm0, %xmm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
+; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x04]
+; CHECK-NEXT:    vpcmpleb %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0x75,0x09,0x3f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+declare i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
+
+define <8 x i16> @test_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
+; CHECK-LABEL: test_ucmp_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
+; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
+; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe1,0x05]
+; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe9,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+define <8 x i16> @test_mask_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
+; CHECK-LABEL: test_mask_ucmp_b_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
+; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
+; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x04]
+; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
+  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
+  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
+  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
+  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
+  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
+  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
+  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
+  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
+  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
+  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
+  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
+  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
+  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
+  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
+  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
+  ret <8 x i16> %vec7
+}
+
+declare i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
+
+define <8 x i8> @test_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
+; CHECK-LABEL: test_cmp_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
+; CHECK-NEXT:    vpcmpgtw %xmm0, %xmm1, %k1 ## encoding: [0x62,0xf1,0x75,0x08,0x65,0xc8]
+; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
+; CHECK-NEXT:    vpcmplew %xmm0, %xmm1, %k4 ## encoding: [0x62,0xf3,0xf5,0x08,0x3f,0xe0,0x02]
+; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf1,0x7d,0x08,0x65,0xe9]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_cmp_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
+; CHECK-NEXT:    vpcmpgtw %xmm0, %xmm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
+; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x04]
+; CHECK-NEXT:    vpcmplew %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x09,0x3f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
+; CHECK-LABEL: test_ucmp_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
+; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
+; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe1,0x05]
+; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe9,0x06]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_ucmp_w_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
+; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
+; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x04]
+; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
+; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
+; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone

llvm/trunk/test/CodeGen/X86/avx512bwvl-intrinsics.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw -mattr=+avx512vl --show-mc-encoding| FileCheck %s
 
-; 256-bit
-
-define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
-; CHECK-LABEL: test_cmp_b_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x00]
-; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
-; CHECK-NEXT:    vpcmpltb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x01]
-; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
-; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    vpcmpunordb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x03]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
-; CHECK-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpnleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x06]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vpcmpordb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x07]
-; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %edi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
-; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
-; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
-; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
-; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
-; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
-  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
-  %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
-  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
-  %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
-  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
-  %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
-  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
-  %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
-  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
-  %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
-  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
-  %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
-  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
-  %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
-  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
-  ret <8 x i32> %vec7
-}
-
-define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
-; CHECK-LABEL: test_mask_cmp_b_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x00]
-; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
-; CHECK-NEXT:    vpcmpltb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x01]
-; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
-; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    vpcmpunordb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x03]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
-; CHECK-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpnleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x06]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vpcmpordb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x07]
-; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %edi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
-; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
-; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
-; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
-; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
-; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
-  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
-  %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
-  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
-  %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
-  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
-  %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
-  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
-  %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
-  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
-  %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
-  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
-  %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
-  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
-  %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
-  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
-  ret <8 x i32> %vec7
-}
-
-declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
-
-define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
-; CHECK-LABEL: test_ucmp_b_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x00]
-; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
-; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
-; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
-; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    vpcmpunordub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x03]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpnequb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
-; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vpcmpordub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x07]
-; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %edi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
-; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
-; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
-; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
-; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
-; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
-  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
-  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
-  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
-  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
-  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
-  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
-  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
-  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
-  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
-  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
-  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
-  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
-  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
-  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
-  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
-  ret <8 x i32> %vec7
-}
-
-define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
-; CHECK-LABEL: test_mask_ucmp_b_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x00]
-; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
-; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
-; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
-; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
-; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
-; CHECK-NEXT:    vpcmpunordub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x03]
-; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
-; CHECK-NEXT:    vpcmpnequb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x04]
-; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
-; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vpcmpordub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x07]
-; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
-; CHECK-NEXT:    vmovd %edi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
-; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
-; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
-; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
-; CHECK-NEXT:    vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8]
-; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
-; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
-; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
-  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
-  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
-  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
-  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
-  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
-  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
-  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
-  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
-  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
-  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
-  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
-  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
-  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
-  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
-  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
-  ret <8 x i32> %vec7
-}
-
-declare i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
-
-define <8 x i16> @test_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
-; CHECK-LABEL: test_cmp_w_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltw %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xc9,0x01]
-; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordw %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltw %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnlew %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordw %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xf9,0x07]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-define <8 x i16> @test_mask_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
-; CHECK-LABEL: test_mask_cmp_w_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltw %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd1,0x01]
-; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordw %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltw %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnlew %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xc9,0x07]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-declare i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
-
-define <8 x i16> @test_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
-; CHECK-LABEL: test_ucmp_w_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunorduw %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequw %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xf1,0x06]
-; CHECK-NEXT:    vpcmporduw %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xf9,0x07]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-define <8 x i16> @test_mask_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
-; CHECK-LABEL: test_mask_ucmp_w_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunorduw %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xf9,0x06]
-; CHECK-NEXT:    vpcmporduw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x07]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-declare i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
-
-; 128-bit
-
-define i16 @test_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b) {
-; CHECK-LABEL: test_pcmpeq_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
-  ret i16 %res
-}
-
-define i16 @test_mask_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
-; CHECK-LABEL: test_mask_pcmpeq_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
-  ret i16 %res
-}
-
-declare i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8>, <16 x i8>, i16)
-
-define i8 @test_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b) {
-; CHECK-LABEL: test_pcmpeq_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
-  ret i8 %res
-}
-
-define i8 @test_mask_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
-; CHECK-LABEL: test_mask_pcmpeq_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
-  ret i8 %res
-}
-
-declare i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16>, <8 x i16>, i8)
-
-define i16 @test_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b) {
-; CHECK-LABEL: test_pcmpgt_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
-  ret i16 %res
-}
-
-define i16 @test_mask_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
-; CHECK-LABEL: test_mask_pcmpgt_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
-  ret i16 %res
-}
-
-declare i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8>, <16 x i8>, i16)
-
-define i8 @test_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b) {
-; CHECK-LABEL: test_pcmpgt_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
-  ret i8 %res
-}
-
-define i8 @test_mask_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
-; CHECK-LABEL: test_mask_pcmpgt_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc1]
-; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
-; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
-  ret i8 %res
-}
-
-declare i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16>, <8 x i16>, i8)
-
-define <8 x i16> @test_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
-; CHECK-LABEL: test_cmp_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltb %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordb %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltb %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleb %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordb %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xf9,0x07]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-define <8 x i16> @test_mask_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
-; CHECK-LABEL: test_mask_cmp_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltb %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordb %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltb %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleb %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordb %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xc9,0x07]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-declare i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
-
-define <8 x i16> @test_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
-; CHECK-LABEL: test_ucmp_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordub %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequb %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordub %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xf9,0x07]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-define <8 x i16> @test_mask_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
-; CHECK-LABEL: test_mask_ucmp_b_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordub %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordub %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x07]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x03]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
-  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
-  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
-  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
-  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
-  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
-  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
-  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
-  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
-  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
-  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
-  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
-  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
-  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
-  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
-  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
-  ret <8 x i16> %vec7
-}
-
-declare i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
-
-define <8 x i8> @test_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: test_cmp_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltw %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xc9,0x01]
-; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordw %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltw %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnlew %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordw %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xf9,0x07]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_cmp_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltw %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd1,0x01]
-; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordw %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltw %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnlew %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xc9,0x07]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
-; CHECK-LABEL: test_ucmp_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunorduw %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequw %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xf1,0x06]
-; CHECK-NEXT:    vpcmporduw %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xf9,0x07]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_ucmp_w_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunorduw %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xf9,0x06]
-; CHECK-NEXT:    vpcmporduw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x07]
-; CHECK-NEXT:    kmovd %k2, %eax ## encoding: [0xc5,0xfb,0x93,0xc2]
-; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovd %k3, %eax ## encoding: [0xc5,0xfb,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovd %k4, %eax ## encoding: [0xc5,0xfb,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovd %k5, %eax ## encoding: [0xc5,0xfb,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovd %k6, %eax ## encoding: [0xc5,0xfb,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovd %k7, %eax ## encoding: [0xc5,0xfb,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovd %k1, %eax ## encoding: [0xc5,0xfb,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
-
 define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: test_mask_packs_epi32_rr_128:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
   ret <8 x i16> %1
 }

llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll

 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vminps %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5d,0xc1]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
   ret <4 x float> %res
 }
 declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)
 
+define <8 x i8> @test_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: test_cmp_d_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
+; CHECK-NEXT:    vpcmpgtd %ymm0, %ymm1, %k1 ## encoding: [0x62,0xf1,0x75,0x28,0x66,0xc8]
+; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd9,0x04]
+; CHECK-NEXT:    vpcmpled %ymm0, %ymm1, %k4 ## encoding: [0x62,0xf3,0x75,0x28,0x1f,0xe0,0x02]
+; CHECK-NEXT:    vpcmpgtd %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf1,0x7d,0x28,0x66,0xe9]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_cmp_d_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
+; CHECK-NEXT:    vpcmpgtd %ymm0, %ymm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x66,0xd0]
+; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x04]
+; CHECK-NEXT:    vpcmpled %ymm0, %ymm1, %k6 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x1f,0xf0,0x02]
+; CHECK-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc9]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: test_ucmp_d_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
+; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc9,0x01]
+; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xd1,0x02]
+; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd9,0x04]
+; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe1,0x05]
+; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe9,0x06]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    kxnorw %k0, %k0, %k0 ## encoding: [0xc5,0xfc,0x46,0xc0]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_ucmp_d_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
+; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x01]
+; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd9,0x02]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x04]
+; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xf1,0x05]
+; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x06]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
+; CHECK-LABEL: test_cmp_q_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x28,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpgtq %ymm0, %ymm1, %k1 ## encoding: [0x62,0xf2,0xf5,0x28,0x37,0xc8]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd1,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    vpcmpleq %ymm0, %ymm1, %k4 ## encoding: [0x62,0xf3,0xf5,0x28,0x1f,0xe0,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0c]
+; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf2,0xfd,0x28,0x37,0xe9]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    movl $15, %eax ## encoding: [0xb8,0x0f,0x00,0x00,0x00]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_cmp_q_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpgtq %ymm0, %ymm1, %k2 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0x37,0xd0]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xd9,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    vpcmpleq %ymm0, %ymm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x1f,0xf0,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0c]
+; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x37,0xf9]
+; CHECK-NEXT:    kshiftlw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0c]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
+; CHECK-LABEL: test_ucmp_q_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x28,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xc9,0x01]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xd1,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xe1,0x05]
+; CHECK-NEXT:    kshiftlw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0c]
+; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xe9,0x06]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    movl $15, %eax ## encoding: [0xb8,0x0f,0x00,0x00,0x00]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_ucmp_q_256:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd1,0x01]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd9,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf1,0x05]
+; CHECK-NEXT:    kshiftlw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0c]
+; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf9,0x06]
+; CHECK-NEXT:    kshiftlw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0c]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: test_cmp_d_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x76,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %k1 ## encoding: [0x62,0xf1,0x75,0x08,0x66,0xc8]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd1,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    vpcmpled %xmm0, %xmm1, %k4 ## encoding: [0x62,0xf3,0x75,0x08,0x1f,0xe0,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0c]
+; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf1,0x7d,0x08,0x66,0xe9]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    movl $15, %eax ## encoding: [0xb8,0x0f,0x00,0x00,0x00]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_cmp_d_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %k2 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x66,0xd0]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xd9,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    vpcmpled %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0x75,0x09,0x1f,0xf0,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0c]
+; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x66,0xf9]
+; CHECK-NEXT:    kshiftlw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0c]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: test_ucmp_d_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x76,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xc9,0x01]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xd1,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xe1,0x05]
+; CHECK-NEXT:    kshiftlw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0c]
+; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xe9,0x06]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    movl $15, %eax ## encoding: [0xb8,0x0f,0x00,0x00,0x00]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_ucmp_d_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd1,0x01]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd9,0x02]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x04]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf1,0x05]
+; CHECK-NEXT:    kshiftlw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0c]
+; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf9,0x06]
+; CHECK-NEXT:    kshiftlw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0c]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: test_cmp_q_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x08,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpgtq %xmm0, %xmm1, %k1 ## encoding: [0x62,0xf2,0xf5,0x08,0x37,0xc8]
+; CHECK-NEXT:    kshiftlw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd1,0x02]
+; CHECK-NEXT:    kshiftlw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd9,0x04]
+; CHECK-NEXT:    kshiftlw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    vpcmpleq %xmm0, %xmm1, %k4 ## encoding: [0x62,0xf3,0xf5,0x08,0x1f,0xe0,0x02]
+; CHECK-NEXT:    kshiftlw $14, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0c]
+; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf2,0xfd,0x08,0x37,0xe9]
+; CHECK-NEXT:    kshiftlw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    movl $3, %eax ## encoding: [0xb8,0x03,0x00,0x00,0x00]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_cmp_q_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpgtq %xmm0, %xmm1, %k2 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0x37,0xd0]
+; CHECK-NEXT:    kshiftlw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xd9,0x02]
+; CHECK-NEXT:    kshiftlw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x04]
+; CHECK-NEXT:    kshiftlw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    vpcmpleq %xmm0, %xmm1, %k6 {%k1} ## encoding: [0x62,0xf3,0xf5,0x09,0x1f,0xf0,0x02]
+; CHECK-NEXT:    kshiftlw $14, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0c]
+; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x37,0xf9]
+; CHECK-NEXT:    kshiftlw $14, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0c]
+; CHECK-NEXT:    kshiftlw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone
+
+define <8 x i8> @test_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
+; CHECK-LABEL: test_ucmp_q_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x08,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xc9,0x01]
+; CHECK-NEXT:    kshiftlw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xd1,0x02]
+; CHECK-NEXT:    kshiftlw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd9,0x04]
+; CHECK-NEXT:    kshiftlw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xe1,0x05]
+; CHECK-NEXT:    kshiftlw $14, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x32,0xe4,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k4, %k4 ## encoding: [0xc4,0xe3,0xf9,0x30,0xe4,0x0c]
+; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xe9,0x06]
+; CHECK-NEXT:    kshiftlw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
+; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
+; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
+; CHECK-NEXT:    movl $3, %eax ## encoding: [0xb8,0x03,0x00,0x00,0x00]
+; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 -1)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+define <8 x i8> @test_mask_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
+; CHECK-LABEL: test_mask_ucmp_q_128:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
+; CHECK-NEXT:    kshiftlw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc0,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k0, %k0 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc0,0x0c]
+; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd1,0x01]
+; CHECK-NEXT:    kshiftlw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k2, %k2 ## encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd9,0x02]
+; CHECK-NEXT:    kshiftlw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x32,0xdb,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k3, %k3 ## encoding: [0xc4,0xe3,0xf9,0x30,0xdb,0x0c]
+; CHECK-NEXT:    kxorw %k0, %k0, %k4 ## encoding: [0xc5,0xfc,0x47,0xe0]
+; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x04]
+; CHECK-NEXT:    kshiftlw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x32,0xed,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k5, %k5 ## encoding: [0xc4,0xe3,0xf9,0x30,0xed,0x0c]
+; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf1,0x05]
+; CHECK-NEXT:    kshiftlw $14, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x32,0xf6,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k6, %k6 ## encoding: [0xc4,0xe3,0xf9,0x30,0xf6,0x0c]
+; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf9,0x06]
+; CHECK-NEXT:    kshiftlw $14, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x32,0xff,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k7, %k7 ## encoding: [0xc4,0xe3,0xf9,0x30,0xff,0x0c]
+; CHECK-NEXT:    kshiftlw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; CHECK-NEXT:    kshiftrw $14, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
+; CHECK-NEXT:    kshiftlw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; CHECK-NEXT:    kshiftrw $12, %k1, %k1 ## encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
+; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
+; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
+; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
+; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
+; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
+; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
+; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
+; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
+; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
+; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
+; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
+; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
+; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
+; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
+; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
+; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
+  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
+  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
+  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
+  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
+  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
+  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
+  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
+  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
+  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
+  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
+  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
+  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
+  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
+  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 %mask)
+  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
+  ret <8 x i8> %vec7
+}
+
+declare i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone

llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512vl --show-mc-encoding| FileCheck %s
 
-; 256-bit
-
-define <8 x i8> @test_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
-; CHECK-LABEL: test_cmp_d_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltd %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xc9,0x01]
-; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordd %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltd %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnled %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordd %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_cmp_d_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltd %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd1,0x01]
-; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordd %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltd %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnled %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordd %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
-; CHECK-LABEL: test_ucmp_d_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequd %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordud %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequd %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordud %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_ucmp_d_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequd %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordud %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequd %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordud %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
-; CHECK-LABEL: test_cmp_q_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltq %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordq %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltq %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleq %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordq %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_cmp_q_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltq %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordq %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltq %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleq %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordq %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
-; CHECK-LABEL: test_ucmp_q_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequq %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunorduq %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequq %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xf1,0x06]
-; CHECK-NEXT:    vpcmporduq %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_ucmp_q_256:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequq %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunorduq %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequq %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf9,0x06]
-; CHECK-NEXT:    vpcmporduq %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone
-
-; 128-bit
-
-define <8 x i8> @test_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: test_cmp_d_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltd %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xc9,0x01]
-; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordd %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltd %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnled %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordd %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_cmp_d_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltd %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xd1,0x01]
-; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordd %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltd %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnled %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordd %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-; CHECK-LABEL: test_ucmp_d_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequd %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordud %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequd %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordud %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_ucmp_d_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequd %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordud %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequd %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordud %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
-; CHECK-LABEL: test_cmp_q_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltq %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunordq %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd9,0x03]
-; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltq %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleq %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xf1,0x06]
-; CHECK-NEXT:    vpcmpordq %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_cmp_q_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltq %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunordq %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe1,0x03]
-; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltq %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleq %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xf9,0x06]
-; CHECK-NEXT:    vpcmpordq %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone
-
-define <8 x i8> @test_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
-; CHECK-LABEL: test_ucmp_q_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpcmpequq %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xc9,0x01]
-; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xd1,0x02]
-; CHECK-NEXT:    vpcmpunorduq %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xd9,0x03]
-; CHECK-NEXT:    vpcmpnequq %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xe1,0x04]
-; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xe9,0x05]
-; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xf1,0x06]
-; CHECK-NEXT:    vpcmporduq %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xf9,0x07]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 -1)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-define <8 x i8> @test_mask_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
-; CHECK-LABEL: test_mask_ucmp_q_128:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; CHECK-NEXT:    vpcmpequq %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xc1,0x00]
-; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd1,0x01]
-; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd9,0x02]
-; CHECK-NEXT:    vpcmpunorduq %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xe1,0x03]
-; CHECK-NEXT:    vpcmpnequq %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xe9,0x04]
-; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf1,0x05]
-; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf9,0x06]
-; CHECK-NEXT:    vpcmporduq %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xc9,0x07]
-; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
-; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
-; CHECK-NEXT:    vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1]
-; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
-; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
-; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
-; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
-; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
-; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x08]
-; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
-; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0a]
-; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
-; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0c]
-; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
-; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x0e]
-; CHECK-NEXT:    retq ## encoding: [0xc3]
-  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
-  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
-  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
-  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
-  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
-  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
-  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
-  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
-  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
-  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
-  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
-  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
-  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
-  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
-  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 7, i8 %mask)
-  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
-  ret <8 x i8> %vec7
-}
-
-declare i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone
-
 define void @compr1(i8* %addr, <8 x double> %data, i8 %mask) {
 ; CHECK-LABEL: compr1:
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
 ; CHECK-NEXT:    vcompresspd %zmm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0xfd,0x49,0x8a,0x07]
 ; CHECK-NEXT:    retq ## encoding: [0xc3]
   call void @llvm.x86.avx512.mask.compress.store.pd.512(i8* %addr, <8 x double> %data, i8 %mask)
   ret void