[clang codegen][AArch64] Use llvm.aarch64.neon.fcvtzs/u where it's necessary

[efriedma]

Summary

fptosi/fptoui have similar, but not identical, semantics. In particular, the behavior on overflow is different.

Fixes https://bugs.llvm.org/show_bug.cgi?id=46844 for 64-bit. (This is missing the corresponding change for 32-bit; the 32-bit equivalent to llvm.aarch64.neon.fcvtzs isn't implemented, as far as I can tell.)

[SjoerdMeijer]

Cheers, LGTM

[azharudd]

Hey Eli,

Looks like this is causing the test-suite build to fail. Can you please take a look.

Refer to http://green.lab.llvm.org/green/job/test-suite-verify-machineinstrs-aarch64-O3/8035/consoleFull http://green.lab.llvm.org/green/job/test-suite-verify-machineinstrs-aarch64-O3/8035/consoleFull.

FAILED: SingleSource/UnitTests/Vector/AArch64/CMakeFiles/aarch64_neon_intrinsics.dir/aarch64_neon_intrinsics.c.o
/Users/buildslave/jenkins/workspace/test-suite-verify-machineinstrs-aarch64-O3/test-suite-build/tools/timeit --summary SingleSource/UnitTests/Vector/AArch64/CMakeFiles/aarch64_neon_intrinsics.dir/aarch64_neon_intrinsics.c.o.time /Users/buildslave/jenkins/workspace/test-suite-verify-machineinstrs-aarch64-O3/compiler/bin/clang -DNDEBUG -B /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin -Wno-unused-command-line-argument -mllvm -verify-machineinstrs -O3 -arch arm64 -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS13.5.sdk -w -Werror=date-time -std=c99 -MD -MT SingleSource/UnitTests/Vector/AArch64/CMakeFiles/aarch64_neon_intrinsics.dir/aarch64_neon_intrinsics.c.o -MF SingleSource/UnitTests/Vector/AArch64/CMakeFiles/aarch64_neon_intrinsics.dir/aarch64_neon_intrinsics.c.o.d -o SingleSource/UnitTests/Vector/AArch64/CMakeFiles/aarch64_neon_intrinsics.dir/aarch64_neon_intrinsics.c.o -c /Users/buildslave/jenkins/workspace/test-suite-verify-machineinstrs-aarch64-O3/test-suite/SingleSource/UnitTests/Vector/AArch64/aarch64_neon_intrinsics.c
fatal error: error in backend: Cannot select: intrinsic %llvm.aarch64.neon.fcvtzs

Thanks
Azhar

[efriedma]

Committed dca23ed; should be fixed now.

-Eli

clang/lib/CodeGen/CGBuiltin.cpp

static const ARMVectorIntrinsicInfo AArch64SIMDIntrinsicMap[] = {
   NEONMAP0(vtst_v),
   NEONMAP0(vtstq_v),
   NEONMAP1(vusdot_v, aarch64_neon_usdot, 0),
   NEONMAP1(vusdotq_v, aarch64_neon_usdot, 0),
   NEONMAP1(vusmmlaq_v, aarch64_neon_usmmla, 0),
 };
 
 static const ARMVectorIntrinsicInfo AArch64SISDIntrinsicMap[] = {
   NEONMAP1(vabdd_f64, aarch64_sisd_fabd, Add1ArgType),
   NEONMAP1(vabds_f32, aarch64_sisd_fabd, Add1ArgType),
   NEONMAP1(vabsd_s64, aarch64_neon_abs, Add1ArgType),
   NEONMAP1(vaddlv_s32, aarch64_neon_saddlv, AddRetType | Add1ArgType),
   NEONMAP1(vaddlv_u32, aarch64_neon_uaddlv, AddRetType | Add1ArgType),
   NEONMAP1(vaddlvq_s32, aarch64_neon_saddlv, AddRetType | Add1ArgType),
   NEONMAP1(vaddlvq_u32, aarch64_neon_uaddlv, AddRetType | Add1ArgType),
   NEONMAP1(vaddv_f32, aarch64_neon_faddv, AddRetType | Add1ArgType),
   NEONMAP1(vaddv_s32, aarch64_neon_saddv, AddRetType | Add1ArgType),
static const ARMVectorIntrinsicInfo AArch64SISDIntrinsicMap[] = {
   NEONMAP1(vcvtad_s64_f64, aarch64_neon_fcvtas, AddRetType | Add1ArgType),
   NEONMAP1(vcvtad_u64_f64, aarch64_neon_fcvtau, AddRetType | Add1ArgType),
   NEONMAP1(vcvtas_s32_f32, aarch64_neon_fcvtas, AddRetType | Add1ArgType),
   NEONMAP1(vcvtas_u32_f32, aarch64_neon_fcvtau, AddRetType | Add1ArgType),
   NEONMAP1(vcvtd_n_f64_s64, aarch64_neon_vcvtfxs2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvtd_n_f64_u64, aarch64_neon_vcvtfxu2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvtd_n_s64_f64, aarch64_neon_vcvtfp2fxs, AddRetType | Add1ArgType),
   NEONMAP1(vcvtd_n_u64_f64, aarch64_neon_vcvtfp2fxu, AddRetType | Add1ArgType),
+  NEONMAP1(vcvtd_s64_f64, aarch64_neon_fcvtzs, AddRetType | Add1ArgType),
+  NEONMAP1(vcvtd_u64_f64, aarch64_neon_fcvtzu, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_bf16_f32, aarch64_neon_bfcvt, 0),
   NEONMAP1(vcvtmd_s64_f64, aarch64_neon_fcvtms, AddRetType | Add1ArgType),
   NEONMAP1(vcvtmd_u64_f64, aarch64_neon_fcvtmu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtms_s32_f32, aarch64_neon_fcvtms, AddRetType | Add1ArgType),
   NEONMAP1(vcvtms_u32_f32, aarch64_neon_fcvtmu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtnd_s64_f64, aarch64_neon_fcvtns, AddRetType | Add1ArgType),
   NEONMAP1(vcvtnd_u64_f64, aarch64_neon_fcvtnu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtns_s32_f32, aarch64_neon_fcvtns, AddRetType | Add1ArgType),
   NEONMAP1(vcvtns_u32_f32, aarch64_neon_fcvtnu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtpd_s64_f64, aarch64_neon_fcvtps, AddRetType | Add1ArgType),
   NEONMAP1(vcvtpd_u64_f64, aarch64_neon_fcvtpu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtps_s32_f32, aarch64_neon_fcvtps, AddRetType | Add1ArgType),
   NEONMAP1(vcvtps_u32_f32, aarch64_neon_fcvtpu, AddRetType | Add1ArgType),
   NEONMAP1(vcvts_n_f32_s32, aarch64_neon_vcvtfxs2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvts_n_f32_u32, aarch64_neon_vcvtfxu2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvts_n_s32_f32, aarch64_neon_vcvtfp2fxs, AddRetType | Add1ArgType),
   NEONMAP1(vcvts_n_u32_f32, aarch64_neon_vcvtfp2fxu, AddRetType | Add1ArgType),
+  NEONMAP1(vcvts_s32_f32, aarch64_neon_fcvtzs, AddRetType | Add1ArgType),
+  NEONMAP1(vcvts_u32_f32, aarch64_neon_fcvtzu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtxd_f32_f64, aarch64_sisd_fcvtxn, 0),
   NEONMAP1(vmaxnmv_f32, aarch64_neon_fmaxnmv, AddRetType | Add1ArgType),
   NEONMAP1(vmaxnmvq_f32, aarch64_neon_fmaxnmv, AddRetType | Add1ArgType),
   NEONMAP1(vmaxnmvq_f64, aarch64_neon_fmaxnmv, AddRetType | Add1ArgType),
   NEONMAP1(vmaxv_f32, aarch64_neon_fmaxv, AddRetType | Add1ArgType),
   NEONMAP1(vmaxv_s32, aarch64_neon_smaxv, AddRetType | Add1ArgType),
   NEONMAP1(vmaxv_u32, aarch64_neon_umaxv, AddRetType | Add1ArgType),
   NEONMAP1(vmaxvq_f32, aarch64_neon_fmaxv, AddRetType | Add1ArgType),
static const ARMVectorIntrinsicInfo AArch64SISDIntrinsicMap[] = {
   NEONMAP1(vcvth_n_f16_s32, aarch64_neon_vcvtfxs2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_f16_s64, aarch64_neon_vcvtfxs2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_f16_u32, aarch64_neon_vcvtfxu2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_f16_u64, aarch64_neon_vcvtfxu2fp, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_s32_f16, aarch64_neon_vcvtfp2fxs, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_s64_f16, aarch64_neon_vcvtfp2fxs, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_u32_f16, aarch64_neon_vcvtfp2fxu, AddRetType | Add1ArgType),
   NEONMAP1(vcvth_n_u64_f16, aarch64_neon_vcvtfp2fxu, AddRetType | Add1ArgType),
+  NEONMAP1(vcvth_s32_f16, aarch64_neon_fcvtzs, AddRetType | Add1ArgType),
+  NEONMAP1(vcvth_s64_f16, aarch64_neon_fcvtzs, AddRetType | Add1ArgType),
+  NEONMAP1(vcvth_u32_f16, aarch64_neon_fcvtzu, AddRetType | Add1ArgType),
+  NEONMAP1(vcvth_u64_f16, aarch64_neon_fcvtzu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtmh_s32_f16, aarch64_neon_fcvtms, AddRetType | Add1ArgType),
   NEONMAP1(vcvtmh_s64_f16, aarch64_neon_fcvtms, AddRetType | Add1ArgType),
   NEONMAP1(vcvtmh_u32_f16, aarch64_neon_fcvtmu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtmh_u64_f16, aarch64_neon_fcvtmu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtnh_s32_f16, aarch64_neon_fcvtns, AddRetType | Add1ArgType),
   NEONMAP1(vcvtnh_s64_f16, aarch64_neon_fcvtns, AddRetType | Add1ArgType),
   NEONMAP1(vcvtnh_u32_f16, aarch64_neon_fcvtnu, AddRetType | Add1ArgType),
   NEONMAP1(vcvtnh_u64_f16, aarch64_neon_fcvtnu, AddRetType | Add1ArgType),
case NEON::BI__builtin_neon_vldrq_p128: {
     return Builder.CreateAlignedLoad(Int128Ty, Ptr,
                                      CharUnits::fromQuantity(16));
   }
   case NEON::BI__builtin_neon_vstrq_p128: {
     llvm::Type *Int128PTy = llvm::Type::getIntNPtrTy(getLLVMContext(), 128);
     Value *Ptr = Builder.CreateBitCast(Ops[0], Int128PTy);
     return Builder.CreateDefaultAlignedStore(EmitScalarExpr(E->getArg(1)), Ptr);
   }
-  case NEON::BI__builtin_neon_vcvts_u32_f32:
-  case NEON::BI__builtin_neon_vcvtd_u64_f64:
-    usgn = true;
-    LLVM_FALLTHROUGH;
-  case NEON::BI__builtin_neon_vcvts_s32_f32:
-  case NEON::BI__builtin_neon_vcvtd_s64_f64: {
-    Ops.push_back(EmitScalarExpr(E->getArg(0)));
-    bool Is64 = Ops[0]->getType()->getPrimitiveSizeInBits() == 64;
-    llvm::Type *InTy = Is64 ? Int64Ty : Int32Ty;
-    llvm::Type *FTy = Is64 ? DoubleTy : FloatTy;
-    Ops[0] = Builder.CreateBitCast(Ops[0], FTy);
-    if (usgn)
-      return Builder.CreateFPToUI(Ops[0], InTy);
-    return Builder.CreateFPToSI(Ops[0], InTy);
-  }
   case NEON::BI__builtin_neon_vcvts_f32_u32:
   case NEON::BI__builtin_neon_vcvtd_f64_u64:
     usgn = true;
     LLVM_FALLTHROUGH;
   case NEON::BI__builtin_neon_vcvts_f32_s32:
   case NEON::BI__builtin_neon_vcvtd_f64_s64: {
     Ops.push_back(EmitScalarExpr(E->getArg(0)));
     bool Is64 = Ops[0]->getType()->getPrimitiveSizeInBits() == 64;
else if (Ops[0]->getType()->getPrimitiveSizeInBits() == 32)
       InTy = Int32Ty;
     else
       InTy = Int16Ty;
     Ops[0] = Builder.CreateBitCast(Ops[0], InTy);
     if (usgn)
       return Builder.CreateUIToFP(Ops[0], FTy);
     return Builder.CreateSIToFP(Ops[0], FTy);
   }
-  case NEON::BI__builtin_neon_vcvth_u16_f16:
-    usgn = true;
-    LLVM_FALLTHROUGH;
-  case NEON::BI__builtin_neon_vcvth_s16_f16: {
-    Ops.push_back(EmitScalarExpr(E->getArg(0)));
-    Ops[0] = Builder.CreateBitCast(Ops[0], HalfTy);
-    if (usgn)
-      return Builder.CreateFPToUI(Ops[0], Int16Ty);
-    return Builder.CreateFPToSI(Ops[0], Int16Ty);
-  }
-  case NEON::BI__builtin_neon_vcvth_u32_f16:
-    usgn = true;
-    LLVM_FALLTHROUGH;
-  case NEON::BI__builtin_neon_vcvth_s32_f16: {
-    Ops.push_back(EmitScalarExpr(E->getArg(0)));
-    Ops[0] = Builder.CreateBitCast(Ops[0], HalfTy);
-    if (usgn)
-      return Builder.CreateFPToUI(Ops[0], Int32Ty);
-    return Builder.CreateFPToSI(Ops[0], Int32Ty);
-  }
-  case NEON::BI__builtin_neon_vcvth_u64_f16:
-    usgn = true;
-    LLVM_FALLTHROUGH;
-  case NEON::BI__builtin_neon_vcvth_s64_f16: {
-    Ops.push_back(EmitScalarExpr(E->getArg(0)));
-    Ops[0] = Builder.CreateBitCast(Ops[0], HalfTy);
-    if (usgn)
-      return Builder.CreateFPToUI(Ops[0], Int64Ty);
-    return Builder.CreateFPToSI(Ops[0], Int64Ty);
-  }
   case NEON::BI__builtin_neon_vcvtah_u16_f16:
   case NEON::BI__builtin_neon_vcvtmh_u16_f16:
   case NEON::BI__builtin_neon_vcvtnh_u16_f16:
   case NEON::BI__builtin_neon_vcvtph_u16_f16:
+  case NEON::BI__builtin_neon_vcvth_u16_f16:
   case NEON::BI__builtin_neon_vcvtah_s16_f16:
   case NEON::BI__builtin_neon_vcvtmh_s16_f16:
   case NEON::BI__builtin_neon_vcvtnh_s16_f16:
-  case NEON::BI__builtin_neon_vcvtph_s16_f16: {
+  case NEON::BI__builtin_neon_vcvtph_s16_f16:
+  case NEON::BI__builtin_neon_vcvth_s16_f16: {
     unsigned Int;
     llvm::Type* InTy = Int32Ty;
     llvm::Type* FTy  = HalfTy;
     llvm::Type *Tys[2] = {InTy, FTy};
     Ops.push_back(EmitScalarExpr(E->getArg(0)));
     switch (BuiltinID) {
     default: llvm_unreachable("missing builtin ID in switch!");
     case NEON::BI__builtin_neon_vcvtah_u16_f16:
       Int = Intrinsic::aarch64_neon_fcvtau; break;
     case NEON::BI__builtin_neon_vcvtmh_u16_f16:
       Int = Intrinsic::aarch64_neon_fcvtmu; break;
     case NEON::BI__builtin_neon_vcvtnh_u16_f16:
       Int = Intrinsic::aarch64_neon_fcvtnu; break;
     case NEON::BI__builtin_neon_vcvtph_u16_f16:
       Int = Intrinsic::aarch64_neon_fcvtpu; break;
+    case NEON::BI__builtin_neon_vcvth_u16_f16:
+      Int = Intrinsic::aarch64_neon_fcvtzu; break;
     case NEON::BI__builtin_neon_vcvtah_s16_f16:
       Int = Intrinsic::aarch64_neon_fcvtas; break;
     case NEON::BI__builtin_neon_vcvtmh_s16_f16:
       Int = Intrinsic::aarch64_neon_fcvtms; break;
     case NEON::BI__builtin_neon_vcvtnh_s16_f16:
       Int = Intrinsic::aarch64_neon_fcvtns; break;
     case NEON::BI__builtin_neon_vcvtph_s16_f16:
       Int = Intrinsic::aarch64_neon_fcvtps; break;
+    case NEON::BI__builtin_neon_vcvth_s16_f16:
+      Int = Intrinsic::aarch64_neon_fcvtzs; break;
     }
     Ops[0] = EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "fcvt");
     return Builder.CreateTrunc(Ops[0], Int16Ty);
   }
   case NEON::BI__builtin_neon_vcaleh_f16:
   case NEON::BI__builtin_neon_vcalth_f16:
   case NEON::BI__builtin_neon_vcageh_f16:
   case NEON::BI__builtin_neon_vcagth_f16: {

   case NEON::BI__builtin_neon_vcvt_s16_v:
   case NEON::BI__builtin_neon_vcvt_u16_v:
   case NEON::BI__builtin_neon_vcvtq_s32_v:
   case NEON::BI__builtin_neon_vcvtq_u32_v:
   case NEON::BI__builtin_neon_vcvtq_s64_v:
   case NEON::BI__builtin_neon_vcvtq_u64_v:
   case NEON::BI__builtin_neon_vcvtq_s16_v:
   case NEON::BI__builtin_neon_vcvtq_u16_v: {
-    Ops[0] = Builder.CreateBitCast(Ops[0], GetFloatNeonType(this, Type));
-    if (usgn)
-      return Builder.CreateFPToUI(Ops[0], Ty);
-    return Builder.CreateFPToSI(Ops[0], Ty);
+    Int = usgn ? Intrinsic::aarch64_neon_fcvtzu : Intrinsic::aarch64_neon_fcvtzs;
+    llvm::Type *Tys[2] = { Ty, GetFloatNeonType(this, Type) };
+    return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "vcvtz");
   }
   case NEON::BI__builtin_neon_vcvta_s16_v:
   case NEON::BI__builtin_neon_vcvta_u16_v:
   case NEON::BI__builtin_neon_vcvta_s32_v:
   case NEON::BI__builtin_neon_vcvtaq_s16_v:
   case NEON::BI__builtin_neon_vcvtaq_s32_v:
   case NEON::BI__builtin_neon_vcvta_u32_v:
   case NEON::BI__builtin_neon_vcvtaq_u16_v:

clang/test/CodeGen/aarch64-neon-fcvt-intrinsics.c

 // CHECK-LABEL: define i64 @test_vcvtpd_u64_f64(double %a) #0 {
 // CHECK:   [[VCVTPD_U64_F64_I:%.*]] = call i64 @llvm.aarch64.neon.fcvtpu.i64.f64(double %a) #2
 // CHECK:   ret i64 [[VCVTPD_U64_F64_I]]
 uint64_t test_vcvtpd_u64_f64(float64_t a) {
   return (uint64_t)vcvtpd_u64_f64(a);
 }
 
 // CHECK-LABEL: define i32 @test_vcvts_s32_f32(float %a) #0 {
-// CHECK:   [[TMP0:%.*]] = fptosi float %a to i32
+// CHECK:   [[TMP0:%.*]] = call i32 @llvm.aarch64.neon.fcvtzs.i32.f32(float %a)
 // CHECK:   ret i32 [[TMP0]]
 int32_t test_vcvts_s32_f32(float32_t a) {
   return (int32_t)vcvts_s32_f32(a);
 }
 
 // CHECK-LABEL: define i64 @test_vcvtd_s64_f64(double %a) #0 {
-// CHECK:   [[TMP0:%.*]] = fptosi double %a to i64
+// CHECK:   [[TMP0:%.*]] = call i64 @llvm.aarch64.neon.fcvtzs.i64.f64(double %a)
 // CHECK:   ret i64 [[TMP0]]
 int64_t test_vcvtd_s64_f64(float64_t a) {
   return (int64_t)vcvtd_s64_f64(a);
 }
 
 // CHECK-LABEL: define i32 @test_vcvts_u32_f32(float %a) #0 {
-// CHECK:   [[TMP0:%.*]] = fptoui float %a to i32
+// CHECK:   [[TMP0:%.*]] = call i32 @llvm.aarch64.neon.fcvtzu.i32.f32(float %a)
 // CHECK:   ret i32 [[TMP0]]
 uint32_t test_vcvts_u32_f32(float32_t a) {
   return (uint32_t)vcvts_u32_f32(a);
 }
 
 // CHECK-LABEL: define i64 @test_vcvtd_u64_f64(double %a) #0 {
-// CHECK:   [[TMP0:%.*]] = fptoui double %a to i64
+// CHECK:   [[TMP0:%.*]] = call i64 @llvm.aarch64.neon.fcvtzu.i64.f64(double %a)
 // CHECK:   ret i64 [[TMP0]]
 uint64_t test_vcvtd_u64_f64(float64_t a) {
   return (uint64_t)vcvtd_u64_f64(a);
 }

clang/test/CodeGen/aarch64-neon-intrinsics.c

 // CHECK:   [[SUB_I:%.*]] = fneg <1 x double> %a
 // CHECK:   ret <1 x double> [[SUB_I]]
 float64x1_t test_vneg_f64(float64x1_t a) {
   return vneg_f64(a);
 }
 
 // CHECK-LABEL: @test_vcvt_s64_f64(
 // CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
-// CHECK:   [[TMP1:%.*]] = fptosi <1 x double> %a to <1 x i64>
+// CHECK:   [[TMP1:%.*]] = call <1 x i64> @llvm.aarch64.neon.fcvtzs.v1i64.v1f64(<1 x double> %a)
 // CHECK:   ret <1 x i64> [[TMP1]]
 int64x1_t test_vcvt_s64_f64(float64x1_t a) {
   return vcvt_s64_f64(a);
 }
 
 // CHECK-LABEL: @test_vcvt_u64_f64(
 // CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
-// CHECK:   [[TMP1:%.*]] = fptoui <1 x double> %a to <1 x i64>
+// CHECK:   [[TMP1:%.*]] = call <1 x i64> @llvm.aarch64.neon.fcvtzu.v1i64.v1f64(<1 x double> %a)
 // CHECK:   ret <1 x i64> [[TMP1]]
 uint64x1_t test_vcvt_u64_f64(float64x1_t a) {
   return vcvt_u64_f64(a);
 }
 
 // CHECK-LABEL: @test_vcvtn_s64_f64(
 // CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
 // CHECK:   [[VCVTN1_I:%.*]] = call <1 x i64> @llvm.aarch64.neon.fcvtns.v1i64.v1f64(<1 x double> %a)

clang/test/CodeGen/aarch64-neon-misc.c

 // CHECK:   [[VRNDI1_I:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> %a)
 // CHECK:   ret <2 x double> [[VRNDI1_I]]
 float64x2_t test_vrndiq_f64(float64x2_t a) {
   return vrndiq_f64(a);
 }
 
 // CHECK-LABEL: @test_vcvt_s32_f32(
 // CHECK:   [[TMP0:%.*]] = bitcast <2 x float> %a to <8 x i8>
-// CHECK:   [[TMP1:%.*]] = fptosi <2 x float> %a to <2 x i32>
+// CHECK:   [[TMP1:%.*]] = call <2 x i32> @llvm.aarch64.neon.fcvtzs.v2i32.v2f32(<2 x float> %a)
 // CHECK:   ret <2 x i32> [[TMP1]]
 int32x2_t test_vcvt_s32_f32(float32x2_t a) {
   return vcvt_s32_f32(a);
 }
 
 // CHECK-LABEL: @test_vcvtq_s32_f32(
 // CHECK:   [[TMP0:%.*]] = bitcast <4 x float> %a to <16 x i8>
-// CHECK:   [[TMP1:%.*]] = fptosi <4 x float> %a to <4 x i32>
+// CHECK:   [[TMP1:%.*]] = call <4 x i32> @llvm.aarch64.neon.fcvtzs.v4i32.v4f32(<4 x float> %a)
 // CHECK:   ret <4 x i32> [[TMP1]]
 int32x4_t test_vcvtq_s32_f32(float32x4_t a) {
   return vcvtq_s32_f32(a);
 }
 
 // CHECK-LABEL: @test_vcvtq_s64_f64(
 // CHECK:   [[TMP0:%.*]] = bitcast <2 x double> %a to <16 x i8>
-// CHECK:   [[TMP1:%.*]] = fptosi <2 x double> %a to <2 x i64>
+// CHECK:   [[TMP1:%.*]] = call <2 x i64> @llvm.aarch64.neon.fcvtzs.v2i64.v2f64(<2 x double> %a)
 // CHECK:   ret <2 x i64> [[TMP1]]
 int64x2_t test_vcvtq_s64_f64(float64x2_t a) {
   return vcvtq_s64_f64(a);
 }
 
 // CHECK-LABEL: @test_vcvt_u32_f32(
 // CHECK:   [[TMP0:%.*]] = bitcast <2 x float> %a to <8 x i8>
-// CHECK:   [[TMP1:%.*]] = fptoui <2 x float> %a to <2 x i32>
+// CHECK:   [[TMP1:%.*]] = call <2 x i32> @llvm.aarch64.neon.fcvtzu.v2i32.v2f32(<2 x float> %a)
 // CHECK:   ret <2 x i32> [[TMP1]]
 uint32x2_t test_vcvt_u32_f32(float32x2_t a) {
   return vcvt_u32_f32(a);
 }
 
 // CHECK-LABEL: @test_vcvtq_u32_f32(
 // CHECK:   [[TMP0:%.*]] = bitcast <4 x float> %a to <16 x i8>
-// CHECK:   [[TMP1:%.*]] = fptoui <4 x float> %a to <4 x i32>
+// CHECK:   [[TMP1:%.*]] = call <4 x i32> @llvm.aarch64.neon.fcvtzu.v4i32.v4f32(<4 x float> %a)
 // CHECK:   ret <4 x i32> [[TMP1]]
 uint32x4_t test_vcvtq_u32_f32(float32x4_t a) {
   return vcvtq_u32_f32(a);
 }
 
 // CHECK-LABEL: @test_vcvtq_u64_f64(
 // CHECK:   [[TMP0:%.*]] = bitcast <2 x double> %a to <16 x i8>
-// CHECK:   [[TMP1:%.*]] = fptoui <2 x double> %a to <2 x i64>
+// CHECK:   [[TMP1:%.*]] = call <2 x i64> @llvm.aarch64.neon.fcvtzu.v2i64.v2f64(<2 x double> %a)
 // CHECK:   ret <2 x i64> [[TMP1]]
 uint64x2_t test_vcvtq_u64_f64(float64x2_t a) {
   return vcvtq_u64_f64(a);
 }
 
 // CHECK-LABEL: @test_vcvtn_s32_f32(
 // CHECK:   [[TMP0:%.*]] = bitcast <2 x float> %a to <8 x i8>
 // CHECK:   [[VCVTN1_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.fcvtns.v2i32.v2f32(<2 x float> %a)

clang/test/CodeGen/aarch64-v8.2a-fp16-intrinsics.c

 // CHECK-LABEL: test_vcvth_f16_u64
 // CHECK:  [[VCVT:%.*]] = uitofp i64 %a to half
 // CHECK:  ret half [[VCVT]]
 float16_t test_vcvth_f16_u64 (uint64_t a) {
   return vcvth_f16_u64(a);
 }
 
 // CHECK-LABEL: test_vcvth_s16_f16
-// CHECK:  [[VCVT:%.*]] = fptosi half %a to i16
-// CHECK:  ret i16 [[VCVT]]
+// CHECK:  [[VCVT:%.*]] = call i32 @llvm.aarch64.neon.fcvtzs.i32.f16(half %a)
+// CHECK:  [[TRUNC:%.*]] = trunc i32 [[VCVT]] to i16
+// CHECK:  ret i16 [[TRUNC]]
 int16_t test_vcvth_s16_f16 (float16_t a) {
   return vcvth_s16_f16(a);
 }
 
 // CHECK-LABEL: test_vcvth_s32_f16
-// CHECK:  [[VCVT:%.*]] = fptosi half %a to i32
+// CHECK:  [[VCVT:%.*]] = call i32 @llvm.aarch64.neon.fcvtzs.i32.f16(half %a)
 // CHECK:  ret i32 [[VCVT]]
 int32_t test_vcvth_s32_f16 (float16_t a) {
   return vcvth_s32_f16(a);
 }
 
 // CHECK-LABEL: test_vcvth_s64_f16
-// CHECK:  [[VCVT:%.*]] = fptosi half %a to i64
+// CHECK:  [[VCVT:%.*]] = call i64 @llvm.aarch64.neon.fcvtzs.i64.f16(half %a)
 // CHECK:  ret i64 [[VCVT]]
 int64_t test_vcvth_s64_f16 (float16_t a) {
   return vcvth_s64_f16(a);
 }
 
 // CHECK-LABEL: test_vcvth_u16_f16
-// CHECK:  [[VCVT:%.*]] = fptoui half %a to i16
-// CHECK:  ret i16 [[VCVT]]
+// CHECK:  [[VCVT:%.*]] = call i32 @llvm.aarch64.neon.fcvtzu.i32.f16(half %a)
+// CHECK:  [[TRUNC:%.*]] = trunc i32 [[VCVT]] to i16
+// CHECK:  ret i16 [[TRUNC]]
 uint16_t test_vcvth_u16_f16 (float16_t a) {
   return vcvth_u16_f16(a);
 }
 
 // CHECK-LABEL: test_vcvth_u32_f16
-// CHECK:  [[VCVT:%.*]] = fptoui half %a to i32
+// CHECK:  [[VCVT:%.*]] = call i32 @llvm.aarch64.neon.fcvtzu.i32.f16(half %a)
 // CHECK:  ret i32 [[VCVT]]
 uint32_t test_vcvth_u32_f16 (float16_t a) {
   return vcvth_u32_f16(a);
 }
 
 // CHECK-LABEL: test_vcvth_u64_f16
-// CHECK:  [[VCVT:%.*]] = fptoui half %a to i64
+// CHECK:  [[VCVT:%.*]] = call i64 @llvm.aarch64.neon.fcvtzu.i64.f16(half %a)
 // CHECK:  ret i64 [[VCVT]]
 uint64_t test_vcvth_u64_f16 (float16_t a) {
   return vcvth_u64_f16(a);
 }
 
 // CHECK-LABEL: test_vcvtah_s16_f16
 // CHECK: [[FCVT:%.*]] = call i32 @llvm.aarch64.neon.fcvtas.i32.f16(half %a)
 // CHECK: [[RET:%.*]] = trunc i32 [[FCVT]] to i16

clang/test/CodeGen/aarch64-v8.2a-neon-intrinsics.c

 // CHECK-LABEL: test_vcvtq_f16_u16
 // CHECK:  [[VCVT:%.*]] = uitofp <8 x i16> %a to <8 x half>
 // CHECK:  ret <8 x half> [[VCVT]]
 float16x8_t test_vcvtq_f16_u16 (uint16x8_t a) {
   return vcvtq_f16_u16(a);
 }
 
 // CHECK-LABEL: test_vcvt_s16_f16
-// CHECK:  [[VCVT:%.*]] = fptosi <4 x half> %a to <4 x i16>
+// CHECK:  [[VCVT:%.*]] = call <4 x i16> @llvm.aarch64.neon.fcvtzs.v4i16.v4f16(<4 x half> %a)
 // CHECK:  ret <4 x i16> [[VCVT]]
 int16x4_t test_vcvt_s16_f16 (float16x4_t a) {
   return vcvt_s16_f16(a);
 }
 
 // CHECK-LABEL: test_vcvtq_s16_f16
-// CHECK:  [[VCVT:%.*]] = fptosi <8 x half> %a to <8 x i16>
+// CHECK:  [[VCVT:%.*]] = call <8 x i16> @llvm.aarch64.neon.fcvtzs.v8i16.v8f16(<8 x half> %a)
 // CHECK:  ret <8 x i16> [[VCVT]]
 int16x8_t test_vcvtq_s16_f16 (float16x8_t a) {
   return vcvtq_s16_f16(a);
 }
 
 // CHECK-LABEL: test_vcvt_u16_f16
-// CHECK:  [[VCVT:%.*]] = fptoui <4 x half> %a to <4 x i16>
+// CHECK:  [[VCVT:%.*]] = call <4 x i16> @llvm.aarch64.neon.fcvtzu.v4i16.v4f16(<4 x half> %a)
 // CHECK:  ret <4 x i16> [[VCVT]]
 uint16x4_t test_vcvt_u16_f16 (float16x4_t a) {
   return vcvt_u16_f16(a);
 }
 
 // CHECK-LABEL: test_vcvtq_u16_f16
-// CHECK:  [[VCVT:%.*]] = fptoui <8 x half> %a to <8 x i16>
+// CHECK:  [[VCVT:%.*]] = call <8 x i16> @llvm.aarch64.neon.fcvtzu.v8i16.v8f16(<8 x half> %a)
 // CHECK:  ret <8 x i16> [[VCVT]]
 uint16x8_t test_vcvtq_u16_f16 (float16x8_t a) {
   return vcvtq_u16_f16(a);
 }
 
 // CHECK-LABEL: test_vcvta_s16_f16
 // CHECK:  [[VCVT:%.*]] = call <4 x i16> @llvm.aarch64.neon.fcvtas.v4i16.v4f16(<4 x half> %a)
 // CHECK:  ret <4 x i16> [[VCVT]]