[X86] Remove GCCBuiltins from 512-bit cvt(u)qqtops, cvt(u)qqtopd, and cvt(u)dqtops intrinsics. Add new variadic uitofp/sitofp with rounding mode intrinsics.

[craig.topper]

Summary

See clang patch D56998 for a full description.

[RKSimon]

include/llvm/IR/IntrinsicsX86.td

2691 ↗

(On Diff #182746)

AFAICT these are only ever going to be used by avx512 targets - add avx512 prefix?

[craig.topper]

Rename the intrinsics and add autoupgrade support

[craig.topper]

Ping

[RKSimon]

LGTM

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

def int_x86_avx512_packuswb_512 : GCCBuiltin<"__builtin_ia32_packuswb512">,
                          [IntrNoMem]>;
   def int_x86_avx512_packusdw_512 : GCCBuiltin<"__builtin_ia32_packusdw512">,
               Intrinsic<[llvm_v32i16_ty], [llvm_v16i32_ty, llvm_v16i32_ty],
                          [IntrNoMem]>;
 }
 
 // Vector convert
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
-  def int_x86_avx512_mask_cvtdq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtdq2ps512_mask">,
-          Intrinsic<[llvm_v16f32_ty],
-          [llvm_v16i32_ty, llvm_v16f32_ty,  llvm_i16_ty,  llvm_i32_ty],
+  def int_x86_avx512_sitofp_round :
+          Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty],
+          [IntrNoMem]>;
+
+  def int_x86_avx512_uitofp_round :
+          Intrinsic<[llvm_anyfloat_ty], [llvm_anyint_ty, llvm_i32_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvtpd2dq_128 :
         GCCBuiltin<"__builtin_ia32_cvtpd2dq128_mask">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v2f64_ty, llvm_v4i32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
def int_x86_avx512_mask_cvtps2uqq_256 :
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvtps2uqq_512 :
         GCCBuiltin<"__builtin_ia32_cvtps2uqq512_mask">,
           Intrinsic<[llvm_v8i64_ty],
           [llvm_v8f32_ty, llvm_v8i64_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtqq2pd_512 :
-        GCCBuiltin<"__builtin_ia32_cvtqq2pd512_mask">,
-          Intrinsic<[llvm_v8f64_ty],
-          [llvm_v8i64_ty, llvm_v8f64_ty,  llvm_i8_ty,  llvm_i32_ty],
-          [IntrNoMem]>;
-
   def int_x86_avx512_mask_cvtqq2ps_128 :
         GCCBuiltin<"__builtin_ia32_cvtqq2ps128_mask">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v2i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtqq2ps_256 :
-        GCCBuiltin<"__builtin_ia32_cvtqq2ps256_mask">,
-          Intrinsic<[llvm_v4f32_ty],
-          [llvm_v4i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
-          [IntrNoMem]>;
-
-  def int_x86_avx512_mask_cvtqq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtqq2ps512_mask">,
-          Intrinsic<[llvm_v8f32_ty],
-          [llvm_v8i64_ty, llvm_v8f32_ty,  llvm_i8_ty,  llvm_i32_ty],
-          [IntrNoMem]>;
-
   def int_x86_avx512_mask_cvttpd2dq_128 :
         GCCBuiltin<"__builtin_ia32_cvttpd2dq128_mask">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v2f64_ty, llvm_v4i32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvttpd2dq_512 :
         GCCBuiltin<"__builtin_ia32_cvttpd2dq512_mask">,
def int_x86_avx512_mask_cvttps2uqq_256 :
           [IntrNoMem]>;
 
   def int_x86_avx512_mask_cvttps2uqq_512 :
         GCCBuiltin<"__builtin_ia32_cvttps2uqq512_mask">,
           Intrinsic<[llvm_v8i64_ty],
           [llvm_v8f32_ty, llvm_v8i64_ty,  llvm_i8_ty,  llvm_i32_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtudq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtudq2ps512_mask">,
-          Intrinsic<[llvm_v16f32_ty],
-          [llvm_v16i32_ty, llvm_v16f32_ty,  llvm_i16_ty,  llvm_i32_ty],
-          [IntrNoMem]>;
-
-  def int_x86_avx512_mask_cvtuqq2pd_512 :
-        GCCBuiltin<"__builtin_ia32_cvtuqq2pd512_mask">,
-          Intrinsic<[llvm_v8f64_ty],
-          [llvm_v8i64_ty, llvm_v8f64_ty,  llvm_i8_ty,  llvm_i32_ty],
-          [IntrNoMem]>;
-
   def int_x86_avx512_mask_cvtuqq2ps_128 :
         GCCBuiltin<"__builtin_ia32_cvtuqq2ps128_mask">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v2i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
           [IntrNoMem]>;
 
-  def int_x86_avx512_mask_cvtuqq2ps_256 :
-        GCCBuiltin<"__builtin_ia32_cvtuqq2ps256_mask">,
-          Intrinsic<[llvm_v4f32_ty],
-          [llvm_v4i64_ty, llvm_v4f32_ty,  llvm_i8_ty],
-          [IntrNoMem]>;
-
-  def int_x86_avx512_mask_cvtuqq2ps_512 :
-        GCCBuiltin<"__builtin_ia32_cvtuqq2ps512_mask">,
-          Intrinsic<[llvm_v8f32_ty],
-          [llvm_v8i64_ty, llvm_v8f32_ty,  llvm_i8_ty,  llvm_i32_ty],
-          [IntrNoMem]>;
-
   def int_x86_avx512_mask_rndscale_pd_128 : GCCBuiltin<"__builtin_ia32_rndscalepd_128_mask">,
         Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty, llvm_i32_ty,
                                      llvm_v2f64_ty,  llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_mask_rndscale_pd_256 : GCCBuiltin<"__builtin_ia32_rndscalepd_256_mask">,
         Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty, llvm_i32_ty,
                                      llvm_v4f64_ty,  llvm_i8_ty], [IntrNoMem]>;
   def int_x86_avx512_mask_rndscale_pd_512 : GCCBuiltin<"__builtin_ia32_rndscalepd_mask">,
         Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_i32_ty, llvm_v8f64_ty,

llvm/trunk/lib/IR/AutoUpgrade.cpp

if (Name == "addcarryx.u32" || // Added in 8.0
       Name.startswith("avx512.mask.andn.") || // Added in 3.9
       Name.startswith("avx512.mask.or.") || // Added in 3.9
       Name.startswith("avx512.mask.xor.") || // Added in 3.9
       Name.startswith("avx512.mask.padd.") || // Added in 4.0
       Name.startswith("avx512.mask.psub.") || // Added in 4.0
       Name.startswith("avx512.mask.pmull.") || // Added in 4.0
       Name.startswith("avx512.mask.cvtdq2pd.") || // Added in 4.0
       Name.startswith("avx512.mask.cvtudq2pd.") || // Added in 4.0
-      Name == "avx512.mask.cvtudq2ps.128" || // Added in 7.0
-      Name == "avx512.mask.cvtudq2ps.256" || // Added in 7.0
-      Name == "avx512.mask.cvtqq2pd.128" || // Added in 7.0
-      Name == "avx512.mask.cvtqq2pd.256" || // Added in 7.0
-      Name == "avx512.mask.cvtuqq2pd.128" || // Added in 7.0
-      Name == "avx512.mask.cvtuqq2pd.256" || // Added in 7.0
-      Name == "avx512.mask.cvtdq2ps.128" || // Added in 7.0
-      Name == "avx512.mask.cvtdq2ps.256" || // Added in 7.0
+      Name.startswith("avx512.mask.cvtudq2ps.") || // Added in 7.0 updated 9.0
+      Name.startswith("avx512.mask.cvtqq2pd.") || // Added in 7.0 updated 9.0
+      Name.startswith("avx512.mask.cvtuqq2pd.") || // Added in 7.0 updated 9.0
+      Name.startswith("avx512.mask.cvtdq2ps.") || // Added in 7.0 updated 9.0
+      Name == "avx512.mask.cvtqq2ps.256" || // Added in 9.0
+      Name == "avx512.mask.cvtqq2ps.512" || // Added in 9.0
+      Name == "avx512.mask.cvtuqq2ps.256" || // Added in 9.0
+      Name == "avx512.mask.cvtuqq2ps.512" || // Added in 9.0
       Name == "avx512.mask.cvtpd2dq.256" || // Added in 7.0
       Name == "avx512.mask.cvtpd2ps.256" || // Added in 7.0
       Name == "avx512.mask.cvttpd2dq.256" || // Added in 7.0
       Name == "avx512.mask.cvttps2dq.128" || // Added in 7.0
       Name == "avx512.mask.cvttps2dq.256" || // Added in 7.0
       Name == "avx512.mask.cvtps2pd.128" || // Added in 7.0
       Name == "avx512.mask.cvtps2pd.256" || // Added in 7.0
       Name == "avx512.cvtusi2sd" || // Added in 7.0
if (IsX86 && (Name.startswith("sse2.pcmp") ||
       Rep = Builder.CreateFPExt(Rep, CI->getType()->getVectorElementType());
       Rep = Builder.CreateInsertElement(CI->getArgOperand(0), Rep, (uint64_t)0);
     } else if (IsX86 && (Name == "sse2.cvtdq2pd" ||
                          Name == "sse2.cvtdq2ps" ||
                          Name == "avx.cvtdq2.pd.256" ||
                          Name == "avx.cvtdq2.ps.256" ||
                          Name.startswith("avx512.mask.cvtdq2pd.") ||
                          Name.startswith("avx512.mask.cvtudq2pd.") ||
-                         Name == "avx512.mask.cvtdq2ps.128" ||
-                         Name == "avx512.mask.cvtdq2ps.256" ||
-                         Name == "avx512.mask.cvtudq2ps.128" ||
-                         Name == "avx512.mask.cvtudq2ps.256" ||
-                         Name == "avx512.mask.cvtqq2pd.128" ||
-                         Name == "avx512.mask.cvtqq2pd.256" ||
-                         Name == "avx512.mask.cvtuqq2pd.128" ||
-                         Name == "avx512.mask.cvtuqq2pd.256" ||
+                         Name.startswith("avx512.mask.cvtdq2ps.") ||
+                         Name.startswith("avx512.mask.cvtudq2ps.") ||
+                         Name.startswith("avx512.mask.cvtqq2pd.") ||
+                         Name.startswith("avx512.mask.cvtuqq2pd.") ||
+                         Name == "avx512.mask.cvtqq2ps.256" ||
+                         Name == "avx512.mask.cvtqq2ps.512" ||
+                         Name == "avx512.mask.cvtuqq2ps.256" ||
+                         Name == "avx512.mask.cvtuqq2ps.512" ||
                          Name == "sse2.cvtps2pd" ||
                          Name == "avx.cvt.ps2.pd.256" ||
                          Name == "avx512.mask.cvtps2pd.128" ||
                          Name == "avx512.mask.cvtps2pd.256")) {
       Type *DstTy = CI->getType();
       Rep = CI->getArgOperand(0);
+      Type *SrcTy = Rep->getType();
 
       unsigned NumDstElts = DstTy->getVectorNumElements();
-      if (NumDstElts < Rep->getType()->getVectorNumElements()) {
+      if (NumDstElts < SrcTy->getVectorNumElements()) {
         assert(NumDstElts == 2 && "Unexpected vector size");
         uint32_t ShuffleMask[2] = { 0, 1 };
         Rep = Builder.CreateShuffleVector(Rep, Rep, ShuffleMask);
       }
 
-      bool IsPS2PD = (StringRef::npos != Name.find("ps2"));
+      bool IsPS2PD = SrcTy->getVectorElementType()->isFloatTy();
       bool IsUnsigned = (StringRef::npos != Name.find("cvtu"));
       if (IsPS2PD)
         Rep = Builder.CreateFPExt(Rep, DstTy, "cvtps2pd");
-      else if (IsUnsigned)
-        Rep = Builder.CreateUIToFP(Rep, DstTy, "cvt");
-      else
-        Rep = Builder.CreateSIToFP(Rep, DstTy, "cvt");
+      else if (CI->getNumArgOperands() == 4 &&
+               (!isa<ConstantInt>(CI->getArgOperand(3)) ||
+                cast<ConstantInt>(CI->getArgOperand(3))->getZExtValue() != 4)) {
+        Intrinsic::ID IID = IsUnsigned ? Intrinsic::x86_avx512_uitofp_round
+                                       : Intrinsic::x86_avx512_sitofp_round;
+        Function *F = Intrinsic::getDeclaration(CI->getModule(), IID,
+                                                { DstTy, SrcTy });
+        Rep = Builder.CreateCall(F, { Rep, CI->getArgOperand(3) });
+      } else {
+        Rep = IsUnsigned ? Builder.CreateUIToFP(Rep, DstTy, "cvt")
+                         : Builder.CreateSIToFP(Rep, DstTy, "cvt");
+      }
 
-      if (CI->getNumArgOperands() == 3)
+      if (CI->getNumArgOperands() >= 3)
         Rep = EmitX86Select(Builder, CI->getArgOperand(2), Rep,
                             CI->getArgOperand(1));
     } else if (IsX86 && (Name.startswith("avx512.mask.loadu."))) {
       Rep = UpgradeMaskedLoad(Builder, CI->getArgOperand(0),
                               CI->getArgOperand(1), CI->getArgOperand(2),
                               /*Aligned*/false);
     } else if (IsX86 && (Name.startswith("avx512.mask.load."))) {
       Rep = UpgradeMaskedLoad(Builder, CI->getArgOperand(0),

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

static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_conflict_d_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CONFLICT, 0),
   X86_INTRINSIC_DATA(avx512_mask_conflict_q_128, INTR_TYPE_1OP_MASK,
                      X86ISD::CONFLICT, 0),
   X86_INTRINSIC_DATA(avx512_mask_conflict_q_256, INTR_TYPE_1OP_MASK,
                      X86ISD::CONFLICT, 0),
   X86_INTRINSIC_DATA(avx512_mask_conflict_q_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CONFLICT, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cvtdq2ps_512, INTR_TYPE_1OP_MASK,
-                     ISD::SINT_TO_FP, X86ISD::SINT_TO_FP_RND), //er
   X86_INTRINSIC_DATA(avx512_mask_cvtpd2dq_128, CVTPD2DQ_MASK,
                      X86ISD::CVTP2SI, X86ISD::MCVTP2SI),
   X86_INTRINSIC_DATA(avx512_mask_cvtpd2dq_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTP2SI, X86ISD::CVTP2SI_RND),
   X86_INTRINSIC_DATA(avx512_mask_cvtpd2ps,     CVTPD2PS_MASK,
                      X86ISD::VFPROUND, X86ISD::VMFPROUND),
   X86_INTRINSIC_DATA(avx512_mask_cvtpd2ps_512, INTR_TYPE_1OP_MASK,
                      X86ISD::VFPROUND, X86ISD::VFPROUND_RND),
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_cvtps2udq_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTP2UI, X86ISD::CVTP2UI_RND),
   X86_INTRINSIC_DATA(avx512_mask_cvtps2uqq_128, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTP2UI, 0),
   X86_INTRINSIC_DATA(avx512_mask_cvtps2uqq_256, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTP2UI, 0),
   X86_INTRINSIC_DATA(avx512_mask_cvtps2uqq_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTP2UI, X86ISD::CVTP2UI_RND),
-  X86_INTRINSIC_DATA(avx512_mask_cvtqq2pd_512, INTR_TYPE_1OP_MASK,
-                     ISD::SINT_TO_FP, X86ISD::SINT_TO_FP_RND),
   X86_INTRINSIC_DATA(avx512_mask_cvtqq2ps_128, CVTQQ2PS_MASK,
                      X86ISD::CVTSI2P, X86ISD::MCVTSI2P),
-  X86_INTRINSIC_DATA(avx512_mask_cvtqq2ps_256, INTR_TYPE_1OP_MASK,
-                     ISD::SINT_TO_FP, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cvtqq2ps_512, INTR_TYPE_1OP_MASK,
-                     ISD::SINT_TO_FP, X86ISD::SINT_TO_FP_RND),
   X86_INTRINSIC_DATA(avx512_mask_cvtsd2ss_round, INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::VFPROUNDS_RND, 0),
   X86_INTRINSIC_DATA(avx512_mask_cvtss2sd_round, INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::VFPEXTS_RND, 0),
   X86_INTRINSIC_DATA(avx512_mask_cvttpd2dq_128, CVTPD2DQ_MASK,
                      X86ISD::CVTTP2SI, X86ISD::MCVTTP2SI),
   X86_INTRINSIC_DATA(avx512_mask_cvttpd2dq_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTTP2SI, X86ISD::CVTTP2SI_RND),
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_mask_cvttps2udq_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTTP2UI, X86ISD::CVTTP2UI_RND),
   X86_INTRINSIC_DATA(avx512_mask_cvttps2uqq_128, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTTP2UI, 0),
   X86_INTRINSIC_DATA(avx512_mask_cvttps2uqq_256, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTTP2UI, 0),
   X86_INTRINSIC_DATA(avx512_mask_cvttps2uqq_512, INTR_TYPE_1OP_MASK,
                      X86ISD::CVTTP2UI, X86ISD::CVTTP2UI_RND),
-  X86_INTRINSIC_DATA(avx512_mask_cvtudq2ps_512, INTR_TYPE_1OP_MASK,
-                     ISD::UINT_TO_FP, X86ISD::UINT_TO_FP_RND),
-  X86_INTRINSIC_DATA(avx512_mask_cvtuqq2pd_512, INTR_TYPE_1OP_MASK,
-                     ISD::UINT_TO_FP, X86ISD::UINT_TO_FP_RND),
   X86_INTRINSIC_DATA(avx512_mask_cvtuqq2ps_128, CVTQQ2PS_MASK,
                      X86ISD::CVTUI2P, X86ISD::MCVTUI2P),
-  X86_INTRINSIC_DATA(avx512_mask_cvtuqq2ps_256, INTR_TYPE_1OP_MASK,
-                     ISD::UINT_TO_FP, 0),
-  X86_INTRINSIC_DATA(avx512_mask_cvtuqq2ps_512, INTR_TYPE_1OP_MASK,
-                     ISD::UINT_TO_FP, X86ISD::UINT_TO_FP_RND),
   X86_INTRINSIC_DATA(avx512_mask_div_sd_round, INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::FDIVS_RND, 0),
   X86_INTRINSIC_DATA(avx512_mask_div_ss_round, INTR_TYPE_SCALAR_MASK_RM,
                      X86ISD::FDIVS_RND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_b_128,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
   X86_INTRINSIC_DATA(avx512_mask_expand_b_256,  COMPRESS_EXPAND_IN_REG,
                      X86ISD::EXPAND, 0),
static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_rsqrt14_ps_256, INTR_TYPE_1OP_MASK, X86ISD::RSQRT14, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt14_ps_512, INTR_TYPE_1OP_MASK, X86ISD::RSQRT14, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt14_sd, INTR_TYPE_SCALAR_MASK, X86ISD::RSQRT14S, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt14_ss, INTR_TYPE_SCALAR_MASK, X86ISD::RSQRT14S, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_pd, INTR_TYPE_1OP_MASK_RM,X86ISD::RSQRT28, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_ps, INTR_TYPE_1OP_MASK_RM,X86ISD::RSQRT28, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_sd, INTR_TYPE_SCALAR_MASK_RM,X86ISD::RSQRT28S, 0),
   X86_INTRINSIC_DATA(avx512_rsqrt28_ss, INTR_TYPE_SCALAR_MASK_RM,X86ISD::RSQRT28S, 0),
+  X86_INTRINSIC_DATA(avx512_sitofp_round, INTR_TYPE_1OP, ISD::SINT_TO_FP, X86ISD::SINT_TO_FP_RND),
   X86_INTRINSIC_DATA(avx512_sqrt_pd_512, INTR_TYPE_1OP, ISD::FSQRT, X86ISD::FSQRT_RND),
   X86_INTRINSIC_DATA(avx512_sqrt_ps_512, INTR_TYPE_1OP, ISD::FSQRT, X86ISD::FSQRT_RND),
   X86_INTRINSIC_DATA(avx512_sub_pd_512, INTR_TYPE_2OP, ISD::FSUB, X86ISD::FSUB_RND),
   X86_INTRINSIC_DATA(avx512_sub_ps_512, INTR_TYPE_2OP, ISD::FSUB, X86ISD::FSUB_RND),
+  X86_INTRINSIC_DATA(avx512_uitofp_round, INTR_TYPE_1OP, ISD::UINT_TO_FP, X86ISD::UINT_TO_FP_RND),
   X86_INTRINSIC_DATA(avx512_vcomi_sd, COMI_RM, X86ISD::COMI, X86ISD::UCOMI),
   X86_INTRINSIC_DATA(avx512_vcomi_ss, COMI_RM, X86ISD::COMI, X86ISD::UCOMI),
   X86_INTRINSIC_DATA(avx512_vcvtsd2si32, INTR_TYPE_1OP, X86ISD::CVTS2SI, X86ISD::CVTS2SI_RND),
   X86_INTRINSIC_DATA(avx512_vcvtsd2si64, INTR_TYPE_1OP, X86ISD::CVTS2SI, X86ISD::CVTS2SI_RND),
   X86_INTRINSIC_DATA(avx512_vcvtsd2usi32, INTR_TYPE_1OP, X86ISD::CVTS2UI, X86ISD::CVTS2UI_RND),
   X86_INTRINSIC_DATA(avx512_vcvtsd2usi64, INTR_TYPE_1OP, X86ISD::CVTS2UI, X86ISD::CVTS2UI_RND),
   X86_INTRINSIC_DATA(avx512_vcvtss2si32, INTR_TYPE_1OP, X86ISD::CVTS2SI, X86ISD::CVTS2SI_RND),
   X86_INTRINSIC_DATA(avx512_vcvtss2si64, INTR_TYPE_1OP, X86ISD::CVTS2SI, X86ISD::CVTS2SI_RND),

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

 ; X64-NEXT:    retq ## encoding: [0xc3]
     %res0 = call <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 -1)
     %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2)
     %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64> %x0, <8 x i32> zeroinitializer, i8 %x2)
     %res3 = add <8 x i32> %res0, %res1
     %res4 = add <8 x i32> %res3, %res2
     ret <8 x i32> %res4
 }
+
+declare <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32>, <16 x float>, i16, i32)
+
+define <16 x float> @test_int_x86_avx512_mask_cvt_dq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_512:
+; X86:       ## %bb.0:
+; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtdq2ps %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0x5b,0xc8]
+; X86-NEXT:    vcvtdq2ps {rn-sae}, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x18,0x5b,0xc0]
+; X86-NEXT:    vaddps %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x74,0x48,0x58,0xc0]
+; X86-NEXT:    retl ## encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_512:
+; X64:       ## %bb.0:
+; X64-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtdq2ps %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x49,0x5b,0xc8]
+; X64-NEXT:    vcvtdq2ps {rn-sae}, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x18,0x5b,0xc0]
+; X64-NEXT:    vaddps %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x74,0x48,0x58,0xc0]
+; X64-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 %x2, i32 4)
+  %res1 = call <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 -1, i32 0)
+  %res2 = fadd <16 x float> %res, %res1
+  ret <16 x float> %res2
+}
+
+declare <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32>, <16 x float>, i16, i32)
+
+define <16 x float> @test_int_x86_avx512_mask_cvt_udq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_512:
+; X86:       ## %bb.0:
+; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 ## encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtudq2ps %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7f,0x49,0x7a,0xc8]
+; X86-NEXT:    vcvtudq2ps {rn-sae}, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7f,0x18,0x7a,0xc0]
+; X86-NEXT:    vaddps %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x74,0x48,0x58,0xc0]
+; X86-NEXT:    retl ## encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_512:
+; X64:       ## %bb.0:
+; X64-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtudq2ps %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7f,0x49,0x7a,0xc8]
+; X64-NEXT:    vcvtudq2ps {rn-sae}, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7f,0x18,0x7a,0xc0]
+; X64-NEXT:    vaddps %zmm0, %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x74,0x48,0x58,0xc0]
+; X64-NEXT:    retq ## encoding: [0xc3]
+  %res = call <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 %x2, i32 4)
+  %res1 = call <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 -1, i32 0)
+  %res2 = fadd <16 x float> %res, %res1
+  ret <16 x float> %res2
+}

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

 ; CHECK-NEXT:    vpmovusdw %zmm0, (%rdi) {%k1}
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
     call void @llvm.x86.avx512.mask.pmovus.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
     call void @llvm.x86.avx512.mask.pmovus.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
     ret void
 }
 
-declare <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32>, <16 x float>, i16, i32)
+declare <16 x float> @llvm.x86.avx512.sitofp.round.v16f32.v16i32(<16 x i32>, i32)
 
 define <16 x float>@test_int_x86_avx512_mask_cvt_dq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_512:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    kmovw %edi, %k1
 ; CHECK-NEXT:    vcvtdq2ps %zmm0, %zmm1 {%k1}
 ; CHECK-NEXT:    vcvtdq2ps {rn-sae}, %zmm0, %zmm0
 ; CHECK-NEXT:    vaddps %zmm0, %zmm1, %zmm0
 ; CHECK-NEXT:    retq
-  %res = call <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 %x2, i32 4)
-  %res1 = call <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 -1, i32 0)
-  %res2 = fadd <16 x float> %res, %res1
+  %cvt = sitofp <16 x i32> %x0 to <16 x float>
+  %1 = bitcast i16 %x2 to <16 x i1>
+  %2 = select <16 x i1> %1, <16 x float> %cvt, <16 x float> %x1
+  %3 = call <16 x float> @llvm.x86.avx512.sitofp.round.v16f32.v16i32(<16 x i32> %x0, i32 0)
+  %res2 = fadd <16 x float> %2, %3
   ret <16 x float> %res2
 }
 
 declare <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double>, <8 x i32>, i8, i32)
 
 define <8 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_512:
 ; CHECK:       ## %bb.0:

 ; CHECK-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
 ; CHECK-NEXT:    retq
   %res = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4)
   %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 8)
   %res2 = add <8 x i32> %res, %res1
   ret <8 x i32> %res2
 }
 
-declare <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32>, <16 x float>, i16, i32)
+declare <16 x float> @llvm.x86.avx512.uitofp.round.v16f32.v16i32(<16 x i32>, i32)
 
 define <16 x float>@test_int_x86_avx512_mask_cvt_udq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_512:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    kmovw %edi, %k1
 ; CHECK-NEXT:    vcvtudq2ps %zmm0, %zmm1 {%k1}
 ; CHECK-NEXT:    vcvtudq2ps {rn-sae}, %zmm0, %zmm0
 ; CHECK-NEXT:    vaddps %zmm0, %zmm1, %zmm0
 ; CHECK-NEXT:    retq
-  %res = call <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 %x2, i32 4)
-  %res1 = call <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 -1, i32 0)
-  %res2 = fadd <16 x float> %res, %res1
+  %cvt = uitofp <16 x i32> %x0 to <16 x float>
+  %1 = bitcast i16 %x2 to <16 x i1>
+  %2 = select <16 x i1> %1, <16 x float> %cvt, <16 x float> %x1
+  %3 = call <16 x float> @llvm.x86.avx512.uitofp.round.v16f32.v16i32(<16 x i32> %x0, i32 0)
+  %res2 = fadd <16 x float> %2, %3
   ret <16 x float> %res2
 }
 
 declare <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double>, <8 x i32>, i8, i32)
 
 define <8 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) {
 ; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_512:
 ; CHECK:       ## %bb.0:

llvm/trunk/test/CodeGen/X86/avx512dq-intrinsics-upgrade.ll

 ; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
 ; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
 ; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
     %res = call i16 @llvm.x86.avx512.mask.fpclass.ps.512(<16 x float> %x0, i32 4, i16 -1)
     %res1 = call i16 @llvm.x86.avx512.mask.fpclass.ps.512(<16 x float> %x0, i32 2, i16 %res)
     ret i16 %res1
 }
+
+declare <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64>, <8 x double>, i8, i32)
+
+define <8 x double>@test_int_x86_avx512_mask_cvt_qq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0xe6,0xc8]
+; X86-NEXT:    vcvtqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0xe6,0xc0]
+; X86-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0xe6,0xc8]
+; X64-NEXT:    vcvtqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0xe6,0xc0]
+; X64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
+; X64-NEXT:    retq # encoding: [0xc3]
+  %res = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4)
+  %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 0)
+  %res2 = fadd <8 x double> %res, %res1
+  ret <8 x double> %res2
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64>, <8 x float>, i8, i32)
+
+define <8 x float> @test_int_x86_avx512_mask_cvt_qq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x49,0x5b,0xc8]
+; X86-NEXT:    vcvtqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xfc,0x18,0x5b,0xc0]
+; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x49,0x5b,0xc8]
+; X64-NEXT:    vcvtqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xfc,0x18,0x5b,0xc0]
+; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
+; X64-NEXT:    retq # encoding: [0xc3]
+  %res = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4)
+  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 0)
+  %res2 = fadd <8 x float> %res, %res1
+  ret <8 x float> %res2
+}
+
+declare <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64>, <8 x double>, i8, i32)
+
+define <8 x double> @test_int_x86_avx512_mask_cvt_uqq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtuqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0x7a,0xc8]
+; X86-NEXT:    vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0x7a,0xc0]
+; X86-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtuqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0x7a,0xc8]
+; X64-NEXT:    vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0x7a,0xc0]
+; X64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
+; X64-NEXT:    retq # encoding: [0xc3]
+  %res = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4)
+  %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 0)
+  %res2 = fadd <8 x double> %res, %res1
+  ret <8 x double> %res2
+}
+
+declare <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64>, <8 x float>, i8, i32)
+
+define <8 x float> @test_int_x86_avx512_mask_cvt_uqq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512:
+; X86:       # %bb.0:
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtuqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x7a,0xc8]
+; X86-NEXT:    vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xff,0x18,0x7a,0xc0]
+; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512:
+; X64:       # %bb.0:
+; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtuqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x7a,0xc8]
+; X64-NEXT:    vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xff,0x18,0x7a,0xc0]
+; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]
+; X64-NEXT:    retq # encoding: [0xc3]
+  %res = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4)
+  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 0)
+  %res2 = fadd <8 x float> %res, %res1
+  ret <8 x float> %res2
+}

llvm/trunk/test/CodeGen/X86/avx512dq-intrinsics.ll

 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.cvtps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 %x2, i32 2)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvtps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 -1, i32 0)
   %res2 = add <8 x i64> %res, %res1
   ret <8 x i64> %res2
 }
 
-declare <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64>, <8 x double>, i8, i32)
+declare <8 x double> @llvm.x86.avx512.sitofp.round.v8f64.v8i64(<8 x i64>, i32)
 
 define <8 x double>@test_int_x86_avx512_mask_cvt_qq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vcvtqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0xe6,0xc8]
 ; X86-NEXT:    vcvtqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0xe6,0xc0]
 ; X86-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vcvtqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0xe6,0xc8]
 ; X64-NEXT:    vcvtqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0xe6,0xc0]
 ; X64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
-  %res = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4)
-  %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 0)
-  %res2 = fadd <8 x double> %res, %res1
+  %cvt = sitofp <8 x i64> %x0 to <8 x double>
+  %1 = bitcast i8 %x2 to <8 x i1>
+  %2 = select <8 x i1> %1, <8 x double> %cvt, <8 x double> %x1
+  %3 = call <8 x double> @llvm.x86.avx512.sitofp.round.v8f64.v8i64(<8 x i64> %x0, i32 0)
+  %res2 = fadd <8 x double> %2, %3
   ret <8 x double> %res2
 }
 
-declare <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64>, <8 x float>, i8, i32)
+declare <8 x float> @llvm.x86.avx512.sitofp.round.v8f32.v8i64(<8 x i64>, i32)
 
 define <8 x float>@test_int_x86_avx512_mask_cvt_qq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) {
 ; X86-AVX512DQ-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512:
 ; X86-AVX512DQ:       # %bb.0:
 ; X86-AVX512DQ-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-AVX512DQ-NEXT:    vcvtqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x49,0x5b,0xc8]
 ; X86-AVX512DQ-NEXT:    vcvtqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xfc,0x18,0x5b,0xc0]
 ; X86-AVX512DQ-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]

 ;
 ; X64-AVX512DQVL-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512:
 ; X64-AVX512DQVL:       # %bb.0:
 ; X64-AVX512DQVL-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-AVX512DQVL-NEXT:    vcvtqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x49,0x5b,0xc8]
 ; X64-AVX512DQVL-NEXT:    vcvtqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xfc,0x18,0x5b,0xc0]
 ; X64-AVX512DQVL-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
 ; X64-AVX512DQVL-NEXT:    retq # encoding: [0xc3]
-  %res = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4)
-  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 0)
-  %res2 = fadd <8 x float> %res, %res1
+  %cvt = sitofp <8 x i64> %x0 to <8 x float>
+  %1 = bitcast i8 %x2 to <8 x i1>
+  %2 = select <8 x i1> %1, <8 x float> %cvt, <8 x float> %x1
+  %3 = call <8 x float> @llvm.x86.avx512.sitofp.round.v8f32.v8i64(<8 x i64> %x0, i32 0)
+  %res2 = fadd <8 x float> %2, %3
   ret <8 x float> %res2
 }
 
 declare <8 x i64> @llvm.x86.avx512.mask.cvttpd2qq.512(<8 x double>, <8 x i64>, i8, i32)
 
 define <8 x i64>@test_int_x86_avx512_mask_cvtt_pd2qq_512(<8 x double> %x0, <8 x i64> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2qq_512:
 ; X86:       # %bb.0:

 ; X64-NEXT:    vpaddq %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0xd4,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <8 x i64> @llvm.x86.avx512.mask.cvttps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 %x2, i32 4)
   %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvttps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 -1, i32 8)
   %res2 = add <8 x i64> %res, %res1
   ret <8 x i64> %res2
 }
 
-declare <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64>, <8 x double>, i8, i32)
+declare <8 x double> @llvm.x86.avx512.uitofp.round.v8f64.v8i64(<8 x i64>, i32)
 
 define <8 x double>@test_int_x86_avx512_mask_cvt_uqq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512:
 ; X86:       # %bb.0:
 ; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vcvtuqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0x7a,0xc8]
 ; X86-NEXT:    vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0x7a,0xc0]
 ; X86-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512:
 ; X64:       # %bb.0:
 ; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vcvtuqq2pd %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0xfe,0x49,0x7a,0xc8]
 ; X64-NEXT:    vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 # encoding: [0x62,0xf1,0xfe,0x18,0x7a,0xc0]
 ; X64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0xf5,0x48,0x58,0xc0]
 ; X64-NEXT:    retq # encoding: [0xc3]
-  %res = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4)
-  %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 0)
-  %res2 = fadd <8 x double> %res, %res1
+  %cvt = uitofp <8 x i64> %x0 to <8 x double>
+  %1 = bitcast i8 %x2 to <8 x i1>
+  %2 = select <8 x i1> %1, <8 x double> %cvt, <8 x double> %x1
+  %3 = call <8 x double> @llvm.x86.avx512.uitofp.round.v8f64.v8i64(<8 x i64> %x0, i32 0)
+  %res2 = fadd <8 x double> %2, %3
   ret <8 x double> %res2
 }
 
-declare <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64>, <8 x float>, i8, i32)
+declare <8 x float> @llvm.x86.avx512.uitofp.round.v8f32.v8i64(<8 x i64>, i32)
 
 define <8 x float>@test_int_x86_avx512_mask_cvt_uqq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) {
 ; X86-AVX512DQ-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512:
 ; X86-AVX512DQ:       # %bb.0:
 ; X86-AVX512DQ-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-AVX512DQ-NEXT:    vcvtuqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x7a,0xc8]
 ; X86-AVX512DQ-NEXT:    vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xff,0x18,0x7a,0xc0]
 ; X86-AVX512DQ-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # encoding: [0xc5,0xf4,0x58,0xc0]

 ;
 ; X64-AVX512DQVL-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512:
 ; X64-AVX512DQVL:       # %bb.0:
 ; X64-AVX512DQVL-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-AVX512DQVL-NEXT:    vcvtuqq2ps %zmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x49,0x7a,0xc8]
 ; X64-AVX512DQVL-NEXT:    vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 # encoding: [0x62,0xf1,0xff,0x18,0x7a,0xc0]
 ; X64-AVX512DQVL-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
 ; X64-AVX512DQVL-NEXT:    retq # encoding: [0xc3]
-  %res = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4)
-  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 0)
-  %res2 = fadd <8 x float> %res, %res1
+  %cvt = uitofp <8 x i64> %x0 to <8 x float>
+  %1 = bitcast i8 %x2 to <8 x i1>
+  %2 = select <8 x i1> %1, <8 x float> %cvt, <8 x float> %x1
+  %3 = call <8 x float> @llvm.x86.avx512.uitofp.round.v8f32.v8i64(<8 x i64> %x0, i32 0)
+  %res2 = fadd <8 x float> %2, %3
   ret <8 x float> %res2
 }
 
 declare <8 x double> @llvm.x86.avx512.mask.reduce.pd.512(<8 x double>, i32, <8 x double>, i8, i32)
 
 define <8 x double>@test_int_x86_avx512_mask_reduce_pd_512(<8 x double> %x0, <8 x double> %x2, i8 %x3) {
 ; X86-LABEL: test_int_x86_avx512_mask_reduce_pd_512:
 ; X86:       # %bb.0:

llvm/trunk/test/CodeGen/X86/avx512dqvl-intrinsics-upgrade.ll

 ; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
 ; CHECK-NEXT:    # kill: def $al killed $al killed $eax
 ; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
   %res = call i8 @llvm.x86.avx512.mask.fpclass.pd.256(<4 x double> %x0, i32 2, i8 -1)
   %res1 = call i8 @llvm.x86.avx512.mask.fpclass.pd.256(<4 x double> %x0, i32 4, i8 %res)
   ret i8 %res1
 }
+
+declare <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64>, <4 x float>, i8)
+
+define <4 x float> @test_int_x86_avx512_mask_cvt_qq2ps_256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_256:
+; X86:       # %bb.0:
+; X86-NEXT:    vcvtqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xfc,0x28,0x5b,0xd0]
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x5b,0xc8]
+; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
+; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_256:
+; X64:       # %bb.0:
+; X64-NEXT:    vcvtqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xfc,0x28,0x5b,0xd0]
+; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x5b,0xc8]
+; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
+; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X64-NEXT:    retq # encoding: [0xc3]
+  %res = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 %x2)
+  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 -1)
+  %res2 = fadd <4 x float> %res, %res1
+  ret <4 x float> %res2
+}
+
+declare <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64>, <4 x float>, i8)
+
+define <4 x float> @test_int_x86_avx512_mask_cvt_uqq2ps_256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) {
+; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_256:
+; X86:       # %bb.0:
+; X86-NEXT:    vcvtuqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xff,0x28,0x7a,0xd0]
+; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
+; X86-NEXT:    vcvtuqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7a,0xc8]
+; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
+; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X86-NEXT:    retl # encoding: [0xc3]
+;
+; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_256:
+; X64:       # %bb.0:
+; X64-NEXT:    vcvtuqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xff,0x28,0x7a,0xd0]
+; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT:    vcvtuqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7a,0xc8]
+; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
+; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
+; X64-NEXT:    retq # encoding: [0xc3]
+  %res = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 %x2)
+  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 -1)
+  %res2 = fadd <4 x float> %res, %res1
+  ret <4 x float> %res2
+}

llvm/trunk/test/CodeGen/X86/avx512dqvl-intrinsics.ll

; X64-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 %x2)
   %res1 = shufflevector <4 x float> %res, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   %res2 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1)
   %res3 = shufflevector <4 x float> %res2, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   %res4 = fadd <4 x float> %res1, %res3
   ret <4 x float> %res4
 }
 
-declare <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64>, <4 x float>, i8)
-
 define <4 x float>@test_int_x86_avx512_mask_cvt_qq2ps_256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_256:
 ; X86:       # %bb.0:
+; X86-NEXT:    vcvtqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xfc,0x28,0x5b,0xd0]
 ; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vcvtqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x5b,0xc8]
-; X86-NEXT:    vcvtqq2ps %ymm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x28,0x5b,0xc0]
-; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
+; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
 ; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_256:
 ; X64:       # %bb.0:
+; X64-NEXT:    vcvtqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xfc,0x28,0x5b,0xd0]
 ; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vcvtqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x5b,0xc8]
-; X64-NEXT:    vcvtqq2ps %ymm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x28,0x5b,0xc0]
-; X64-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
+; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
 ; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X64-NEXT:    retq # encoding: [0xc3]
-  %res = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 %x2)
-  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 -1)
-  %res2 = fadd <4 x float> %res, %res1
+  %cvt1 = sitofp <4 x i64> %x0 to <4 x float>
+  %1 = bitcast i8 %x2 to <8 x i1>
+  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %2 = select <4 x i1> %extract, <4 x float> %cvt1, <4 x float> %x1
+  %cvt = sitofp <4 x i64> %x0 to <4 x float>
+  %res2 = fadd <4 x float> %2, %cvt
   ret <4 x float> %res2
 }
 
 declare <2 x i64> @llvm.x86.avx512.mask.cvttpd2qq.128(<2 x double>, <2 x i64>, i8)
 
 define <2 x i64>@test_int_x86_avx512_mask_cvtt_pd2qq_128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2qq_128:
 ; X86:       # %bb.0:
; X64-NEXT:    retq # encoding: [0xc3]
   ret <4 x float> %res4
 }
 
 declare <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64>, <4 x float>, i8)
 
 define <4 x float>@test_int_x86_avx512_mask_cvt_uqq2ps_256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_256:
 ; X86:       # %bb.0:
+; X86-NEXT:    vcvtuqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xff,0x28,0x7a,0xd0]
 ; X86-NEXT:    kmovb {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf9,0x90,0x4c,0x24,0x04]
 ; X86-NEXT:    vcvtuqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7a,0xc8]
-; X86-NEXT:    vcvtuqq2ps %ymm0, %xmm0 # encoding: [0x62,0xf1,0xff,0x28,0x7a,0xc0]
-; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
+; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
 ; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X86-NEXT:    retl # encoding: [0xc3]
 ;
 ; X64-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_256:
 ; X64:       # %bb.0:
+; X64-NEXT:    vcvtuqq2ps %ymm0, %xmm2 # encoding: [0x62,0xf1,0xff,0x28,0x7a,0xd0]
 ; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
 ; X64-NEXT:    vcvtuqq2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7a,0xc8]
-; X64-NEXT:    vcvtuqq2ps %ymm0, %xmm0 # encoding: [0x62,0xf1,0xff,0x28,0x7a,0xc0]
-; X64-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
+; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
 ; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
 ; X64-NEXT:    retq # encoding: [0xc3]
-  %res = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 %x2)
-  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 -1)
-  %res2 = fadd <4 x float> %res, %res1
+  %cvt1 = uitofp <4 x i64> %x0 to <4 x float>
+  %1 = bitcast i8 %x2 to <8 x i1>
+  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %2 = select <4 x i1> %extract, <4 x float> %cvt1, <4 x float> %x1
+  %cvt = uitofp <4 x i64> %x0 to <4 x float>
+  %res2 = fadd <4 x float> %2, %cvt
   ret <4 x float> %res2
 }
 
 declare <2 x i64> @llvm.x86.avx512.mask.cvttps2uqq.128(<4 x float>, <2 x i64>, i8)
 
 define <2 x i64>@test_int_x86_avx512_mask_cvtt_ps2uqq_128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) {
 ; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2uqq_128:
 ; X86:       # %bb.0: