[SveEmitter] Add builtins for FP conversions

[sdesmalen]

Summary

This adds the flag IsOverloadCvt which tells CGBulitin to use
the result type and the type of the last operand as the
overloaded types for the LLVM IR intrinsic.

This also adds the flag IsFPConvert, which is needed to avoid
converting the predicate of the operation from svbool_t to
a predicate with fewer lanes, as the LLVM IR intrinsics use
the <vscale x 16 x i1> as the predicate.

[efriedma]

as the LLVM IR intrinsics use the <vscale x 16 x i1> as the predicate.

Why are the fp conversion intrinsics special here? Should we fix the LLVM intrinsic definitions?

[sdesmalen]

In D78239#1985102, @efriedma wrote:

as the LLVM IR intrinsics use the <vscale x 16 x i1> as the predicate.

Why are the fp conversion intrinsics special here? Should we fix the LLVM intrinsic definitions?

I questioned that as well, but didn't really want to complicate this patch and also wasn't sure whether it is worth it because the convert intrinsics are really only ever used by the ACLE.
On the other hand, it would let us get rid of the IsFPConvert flag and the definition would better match the behaviour of the instruction.

I'm happy to change their definitions and remove the IsFPConvert flag if that has your preference.

[efriedma]

I think the predicate type is something we should fix. Even if it's not causing problems now, it seems like the sort of thing that will bite us later.

It doesn't necessarily need to block this patch, I guess, but it at least needs a good comment in EmitAArch64SVEBuiltinExpr explaining what's happening.

clang/utils/TableGen/SveEmitter.cpp

1045

Not worth implementing these separately? I guess I can see why it wouldn't be profitable.

[sdesmalen]

In D78239#1987859, @efriedma wrote:

I think the predicate type is something we should fix. Even if it's not causing problems now, it seems like the sort of thing that will bite us later.

Yes I agree, I've implemented this in D78402

clang/utils/TableGen/SveEmitter.cpp

1045

We currently don't really optimise the _x (MergeAny) case yet for other intrinsics, so addressing this seems more trouble than it's worth, specifically because the merging with undef for only the top/odd lanes would not be trivial to represent.

[sdesmalen]

• Added comment to CGBuiltin explaining that the special case for IsFPConvert will be removed in a follow-up patch.

[efriedma]

LGTM

clang/utils/TableGen/SveEmitter.cpp

1045

Okay.

clang/include/clang/Basic/TargetBuiltins.h

#undef LLVM_GET_SVE_IMMCHECKTYPES
     bool isStructLoad() const { return Flags & IsStructLoad; }
     bool isStructStore() const { return Flags & IsStructStore; }
     bool isZExtReturn() const { return Flags & IsZExtReturn; }
     bool isByteIndexed() const { return Flags & IsByteIndexed; }
     bool isOverloadNone() const { return Flags & IsOverloadNone; }
     bool isOverloadWhile() const { return Flags & IsOverloadWhile; }
     bool isOverloadDefault() const { return !(Flags & OverloadKindMask); }
     bool isOverloadWhileRW() const { return Flags & IsOverloadWhileRW; }
+    bool isOverloadCvt() const { return Flags & IsOverloadCvt; }
+    bool isFPConvert() const { return Flags & IsFPConvert; }
 
     uint64_t getBits() const { return Flags; }
     bool isFlagSet(uint64_t Flag) const { return Flags & Flag; }
   };
 
   /// Hexagon builtins
   namespace Hexagon {
     enum {

clang/include/clang/Basic/arm_sve.td

 def IsGatherLoad              : FlagType<0x00008000>;
 def IsScatterStore            : FlagType<0x00010000>;
 def IsStructLoad              : FlagType<0x00020000>;
 def IsStructStore             : FlagType<0x00040000>;
 def IsZExtReturn              : FlagType<0x00080000>; // Return value is sign-extend by default
 def IsOverloadNone            : FlagType<0x00100000>; // Intrinsic does not take any overloaded types.
 def IsOverloadWhile           : FlagType<0x00200000>; // Use {default type, typeof(operand1)} as overloaded types.
 def IsOverloadWhileRW         : FlagType<0x00400000>; // Use {pred(default type), typeof(operand0)} as overloaded types.
+def IsOverloadCvt             : FlagType<0x00800000>; // Use {typeof(operand0), typeof(last operand)} as overloaded types.
 def OverloadKindMask          : FlagType<0x00E00000>; // When the masked values are all '0', the default type is used as overload type.
-//      :                                     :
-//      :                                     :
-def IsByteIndexed             : FlagType<0x02000000>;
+def IsByteIndexed             : FlagType<0x01000000>;
+def IsFPConvert               : FlagType<0x02000000>;
 
 // These must be kept in sync with the flags in include/clang/Basic/TargetBuiltins.h
 class ImmCheckType<int val> {
   int Value = val;
 }
 def ImmCheck0_31                : ImmCheckType<0>;  // 0..31 (used for e.g. predicate patterns)
 def ImmCheck1_16                : ImmCheckType<1>;  // 1..16
 def ImmCheckExtract             : ImmCheckType<2>;  // 0..(2048/sizeinbits(elt) - 1)

 defm SVABS_F : SInstZPZ<"svabs", "hfd", "aarch64_sve_fabs">;
 defm SVNEG_F : SInstZPZ<"svneg", "hfd", "aarch64_sve_fneg">;
 
 def SVTMAD  : SInst<"svtmad[_{d}]",  "dddi", "hfd", MergeNone, "aarch64_sve_ftmad_x", [], [ImmCheck<2, ImmCheck0_7>]>;
 
 def SVMLA_LANE  : SInst<"svmla_lane[_{d}]",  "ddddi",  "hfd", MergeNone, "aarch64_sve_fmla_lane", [], [ImmCheck<3, ImmCheckLaneIndex, 2>]>;
 def SVCMLA_LANE : SInst<"svcmla_lane[_{d}]", "ddddii", "hf",  MergeNone, "aarch64_sve_fcmla_lane", [], [ImmCheck<3, ImmCheckLaneIndexCompRotate, 2>,
                                                                                                         ImmCheck<4, ImmCheckComplexRotAll90>]>;
+////////////////////////////////////////////////////////////////////////////////
+// Floating-point conversions
+
+multiclass SInstCvtMXZ<
+    string name, string m_types, string xz_types, string types,
+    string intrinsic, list<FlagType> flags = [IsFPConvert, IsOverloadNone]> {
+  def _M : SInst<name, m_types,  types, MergeOp1,     intrinsic, flags>;
+  def _X : SInst<name, xz_types, types, MergeAnyExp,  intrinsic, flags>;
+  def _Z : SInst<name, xz_types, types, MergeZeroExp, intrinsic, flags>;
+}
+
+multiclass SInstCvtMX<string name, string m_types, string xz_types,
+                      string types, string intrinsic,
+                      list<FlagType> flags = [IsFPConvert, IsOverloadNone]> {
+  def _M : SInst<name, m_types,  types, MergeOp1,     intrinsic, flags>;
+  def _X : SInst<name, xz_types, types, MergeAnyExp,  intrinsic, flags>;
+}
+
+// svcvt_s##_f16
+defm SVFCVTZS_S16_F16 : SInstCvtMXZ<"svcvt_s16[_f16]", "ddPO", "dPO", "s",  "aarch64_sve_fcvtzs", [IsOverloadCvt]>;
+defm SVFCVTZS_S32_F16 : SInstCvtMXZ<"svcvt_s32[_f16]", "ddPO", "dPO", "i",  "aarch64_sve_fcvtzs_i32f16">;
+defm SVFCVTZS_S64_F16 : SInstCvtMXZ<"svcvt_s64[_f16]", "ddPO", "dPO", "l",  "aarch64_sve_fcvtzs_i64f16">;
+
+// svcvt_s##_f32
+defm SVFCVTZS_S32_F32 : SInstCvtMXZ<"svcvt_s32[_f32]", "ddPM", "dPM", "i",  "aarch64_sve_fcvtzs", [IsOverloadCvt]>;
+defm SVFCVTZS_S64_F32 : SInstCvtMXZ<"svcvt_s64[_f32]", "ddPM", "dPM", "l",  "aarch64_sve_fcvtzs_i64f32">;
+
+// svcvt_s##_f64
+defm SVFCVTZS_S32_F64 : SInstCvtMXZ<"svcvt_s32[_f64]", "ddPN", "dPN", "i",  "aarch64_sve_fcvtzs_i32f64">;
+defm SVFCVTZS_S64_F64 : SInstCvtMXZ<"svcvt_s64[_f64]", "ddPN", "dPN", "l",  "aarch64_sve_fcvtzs", [IsOverloadCvt]>;
+
+// svcvt_u##_f16
+defm SVFCVTZU_U16_F16 : SInstCvtMXZ<"svcvt_u16[_f16]", "ddPO", "dPO", "Us", "aarch64_sve_fcvtzu", [IsOverloadCvt]>;
+defm SVFCVTZU_U32_F16 : SInstCvtMXZ<"svcvt_u32[_f16]", "ddPO", "dPO", "Ui", "aarch64_sve_fcvtzu_i32f16">;
+defm SVFCVTZU_U64_F16 : SInstCvtMXZ<"svcvt_u64[_f16]", "ddPO", "dPO", "Ul", "aarch64_sve_fcvtzu_i64f16">;
+
+// svcvt_u##_f32
+defm SVFCVTZU_U32_F32 : SInstCvtMXZ<"svcvt_u32[_f32]", "ddPM", "dPM", "Ui", "aarch64_sve_fcvtzu", [IsOverloadCvt]>;
+defm SVFCVTZU_U64_F32 : SInstCvtMXZ<"svcvt_u64[_f32]", "ddPM", "dPM", "Ul", "aarch64_sve_fcvtzu_i64f32">;
+
+// svcvt_u##_f64
+defm SVFCVTZU_U32_F64 : SInstCvtMXZ<"svcvt_u32[_f64]", "ddPN", "dPN", "Ui", "aarch64_sve_fcvtzu_i32f64">;
+defm SVFCVTZU_U64_F64 : SInstCvtMXZ<"svcvt_u64[_f64]", "ddPN", "dPN", "Ul", "aarch64_sve_fcvtzu", [IsOverloadCvt]>;
+
+// svcvt_f16_s##
+defm SVFCVTZS_F16_S16 : SInstCvtMXZ<"svcvt_f16[_s16]", "OOPd", "OPd", "s",  "aarch64_sve_scvtf", [IsOverloadCvt]>;
+defm SVFCVTZS_F16_S32 : SInstCvtMXZ<"svcvt_f16[_s32]", "OOPd", "OPd", "i",  "aarch64_sve_scvtf_f16i32">;
+defm SVFCVTZS_F16_S64 : SInstCvtMXZ<"svcvt_f16[_s64]", "OOPd", "OPd", "l",  "aarch64_sve_scvtf_f16i64">;
+
+// svcvt_f32_s##
+defm SVFCVTZS_F32_S32 : SInstCvtMXZ<"svcvt_f32[_s32]", "MMPd", "MPd", "i",  "aarch64_sve_scvtf", [IsOverloadCvt]>;
+defm SVFCVTZS_F32_S64 : SInstCvtMXZ<"svcvt_f32[_s64]", "MMPd", "MPd", "l",  "aarch64_sve_scvtf_f32i64">;
+
+// svcvt_f64_s##
+defm SVFCVTZS_F64_S32 : SInstCvtMXZ<"svcvt_f64[_s32]", "NNPd", "NPd", "i",  "aarch64_sve_scvtf_f64i32">;
+defm SVFCVTZS_F64_S64 : SInstCvtMXZ<"svcvt_f64[_s64]", "NNPd", "NPd", "l",  "aarch64_sve_scvtf", [IsOverloadCvt]>;
+
+// svcvt_f16_u##
+defm SVFCVTZU_F16_U16 : SInstCvtMXZ<"svcvt_f16[_u16]", "OOPd", "OPd", "Us", "aarch64_sve_ucvtf", [IsOverloadCvt]>;
+defm SVFCVTZU_F16_U32 : SInstCvtMXZ<"svcvt_f16[_u32]", "OOPd", "OPd", "Ui", "aarch64_sve_ucvtf_f16i32">;
+defm SVFCVTZU_F16_U64 : SInstCvtMXZ<"svcvt_f16[_u64]", "OOPd", "OPd", "Ul", "aarch64_sve_ucvtf_f16i64">;
+
+// svcvt_f32_u##
+defm SVFCVTZU_F32_U32 : SInstCvtMXZ<"svcvt_f32[_u32]", "MMPd", "MPd", "Ui", "aarch64_sve_ucvtf", [IsOverloadCvt]>;
+defm SVFCVTZU_F32_U64 : SInstCvtMXZ<"svcvt_f32[_u64]", "MMPd", "MPd", "Ul", "aarch64_sve_ucvtf_f32i64">;
+
+// svcvt_f64_u##
+defm SVFCVTZU_F64_U32 : SInstCvtMXZ<"svcvt_f64[_u32]", "NNPd", "NPd", "Ui", "aarch64_sve_ucvtf_f64i32">;
+defm SVFCVTZU_F64_U64 : SInstCvtMXZ<"svcvt_f64[_u64]", "NNPd", "NPd", "Ul", "aarch64_sve_ucvtf", [IsOverloadCvt]>;
+
+// svcvt_f16_f##
+defm SVFCVT_F16_F32   : SInstCvtMXZ<"svcvt_f16[_f32]", "OOPd", "OPd", "f", "aarch64_sve_fcvt_f16f32">;
+defm SVFCVT_F16_F64   : SInstCvtMXZ<"svcvt_f16[_f64]", "OOPd", "OPd", "d", "aarch64_sve_fcvt_f16f64">;
+
+// svcvt_f32_f##
+defm SVFCVT_F32_F16   : SInstCvtMXZ<"svcvt_f32[_f16]", "MMPd", "MPd", "h", "aarch64_sve_fcvt_f32f16">;
+defm SVFCVT_F32_F64   : SInstCvtMXZ<"svcvt_f32[_f64]", "MMPd", "MPd", "d", "aarch64_sve_fcvt_f32f64">;
+
+// svcvt_f64_f##
+defm SVFCVT_F64_F16   : SInstCvtMXZ<"svcvt_f64[_f16]", "NNPd", "NPd", "h", "aarch64_sve_fcvt_f64f16">;
+defm SVFCVT_F64_F32   : SInstCvtMXZ<"svcvt_f64[_f32]", "NNPd", "NPd", "f", "aarch64_sve_fcvt_f64f32">;
+
+let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
+defm SVCVTLT_F32    : SInstCvtMX<"svcvtlt_f32[_f16]",  "ddPh", "dPh", "f", "aarch64_sve_fcvtlt_f32f16">;
+defm SVCVTLT_F64    : SInstCvtMX<"svcvtlt_f64[_f32]",  "ddPh", "dPh", "d", "aarch64_sve_fcvtlt_f64f32">;
+
+defm SVCVTX_F32     : SInstCvtMXZ<"svcvtx_f32[_f64]",  "MMPd", "MPd", "d", "aarch64_sve_fcvtx_f32f64">;
+
+def SVCVTNT_F32     : SInst<"svcvtnt_f16[_f32]",  "hhPd", "f", MergeOp1, "aarch64_sve_fcvtnt_f16f32">;
+def SVCVTNT_F64     : SInst<"svcvtnt_f32[_f64]",  "hhPd", "d", MergeOp1, "aarch64_sve_fcvtnt_f32f64">;
+//  SVCVTNT_X       : Implemented as macro by SveEmitter.cpp
+
+def SVCVTXNT_F32    : SInst<"svcvtxnt_f32[_f64]", "MMPd", "d", MergeOp1, "aarch64_sve_fcvtxnt_f32f64">;
+//  SVCVTXNT_X_F32  : Implemented as macro by SveEmitter.cpp
+
+}
 
 def SVCADD_M : SInst<"svcadd[_{d}]", "dPddi",  "hfd", MergeOp1,  "aarch64_sve_fcadd", [], [ImmCheck<3, ImmCheckComplexRot90_270>]>;
 def SVCMLA_M : SInst<"svcmla[_{d}]", "dPdddi", "hfd", MergeOp1,  "aarch64_sve_fcmla", [], [ImmCheck<4, ImmCheckComplexRotAll90>]>;
 
 ////////////////////////////////////////////////////////////////////////////////
 // Saturating scalar arithmetic
 def SVQDECH_S : SInst<"svqdech_pat[_{d}]",   "ddIi", "s", MergeNone, "aarch64_sve_sqdech", [], [ImmCheck<2, ImmCheck1_16>]>;
 def SVQDECH_U : SInst<"svqdech_pat[_{d}]",   "ddIi", "Us", MergeNone, "aarch64_sve_uqdech", [], [ImmCheck<2, ImmCheck1_16>]>;

clang/lib/CodeGen/CGBuiltin.cpp

CodeGenFunction::getSVEOverloadTypes(SVETypeFlags TypeFlags,
   llvm::Type *DefaultType = getSVEType(TypeFlags);
 
   if (TypeFlags.isOverloadWhile())
     return {DefaultType, Ops[1]->getType()};
 
   if (TypeFlags.isOverloadWhileRW())
     return {getSVEPredType(TypeFlags), Ops[0]->getType()};
 
+  if (TypeFlags.isOverloadCvt())
+    return {Ops[0]->getType(), Ops.back()->getType()};
+
   assert(TypeFlags.isOverloadDefault() && "Unexpected value for overloads");
   return {DefaultType};
 }
 
 Value *CodeGenFunction::EmitAArch64SVEBuiltinExpr(unsigned BuiltinID,
                                                   const CallExpr *E) {
   // Find out if any arguments are required to be integer constant expressions.
   unsigned ICEArguments = 0;
if (TypeFlags.getMergeType() == SVETypeFlags::MergeZeroExp)
       InsertExplicitZeroOperand(Builder, Ty, Ops);
 
     if (TypeFlags.getMergeType() == SVETypeFlags::MergeAnyExp)
       InsertExplicitUndefOperand(Builder, Ty, Ops);
 
     // Predicates must match the main datatype.
     for (unsigned i = 0, e = Ops.size(); i != e; ++i)
       if (auto PredTy = dyn_cast<llvm::VectorType>(Ops[i]->getType()))
-        if (PredTy->getElementType()->isIntegerTy(1))
-          Ops[i] = EmitSVEPredicateCast(Ops[i], getSVEType(TypeFlags));
+        if (PredTy->getElementType()->isIntegerTy(1)) {
+          // The special case for `isFPConvert` is because the predicates of the
+          // ACLE IR intrinsics for FP converts are always of type <vscale x 16 x i1>.
+          // This special-case will be removed in a follow-up patch that updates
+          // the FP conversion intrinsics with predicates that match the
+          // default type.
+          llvm::VectorType *NewPredTy =
+              TypeFlags.isFPConvert()
+                  ? llvm::VectorType::get(Builder.getInt1Ty(), {16, true})
+                  : getSVEType(TypeFlags);
+          Ops[i] = EmitSVEPredicateCast(Ops[i], NewPredTy);
+        }
 
     // Splat scalar operand to vector (intrinsics with _n infix)
     if (TypeFlags.hasSplatOperand()) {
       unsigned OpNo = TypeFlags.getSplatOperand();
       Ops[OpNo] = EmitSVEDupX(Ops[OpNo]);
     }
 
     // Predicated intrinsics with _z suffix need a select w/ zeroinitializer.

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_cvt.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svint16_t test_svcvt_s16_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s16_f16_z
+  // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.fcvtzs.nxv8i16.nxv8f16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s16,_f16,_z,)(pg, op);
+}
+
+svint16_t test_svcvt_s16_f16_m(svint16_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s16_f16_m
+  // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.fcvtzs.nxv8i16.nxv8f16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s16,_f16,_m,)(inactive, pg, op);
+}
+
+svint16_t test_svcvt_s16_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s16_f16_x
+  // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.fcvtzs.nxv8i16.nxv8f16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s16,_f16,_x,)(pg, op);
+}
+
+svuint16_t test_svcvt_u16_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u16_f16_z
+  // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.fcvtzu.nxv8i16.nxv8f16(<vscale x 8 x i16> zeroinitializer, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u16,_f16,_z,)(pg, op);
+}
+
+svuint16_t test_svcvt_u16_f16_m(svuint16_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u16_f16_m
+  // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.fcvtzu.nxv8i16.nxv8f16(<vscale x 8 x i16> %inactive, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u16,_f16,_m,)(inactive, pg, op);
+}
+
+svuint16_t test_svcvt_u16_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u16_f16_x
+  // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.fcvtzu.nxv8i16.nxv8f16(<vscale x 8 x i16> undef, <vscale x 8 x i1> %[[PG]], <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u16,_f16,_x,)(pg, op);
+}
+
+svint32_t test_svcvt_s32_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f16_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.i32f16(<vscale x 4 x i32> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f16,_z,)(pg, op);
+}
+
+svint32_t test_svcvt_s32_f32_z(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f32_z
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.nxv4i32.nxv4f32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f32,_z,)(pg, op);
+}
+
+svint32_t test_svcvt_s32_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.i32f64(<vscale x 4 x i32> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f64,_z,)(pg, op);
+}
+
+svint32_t test_svcvt_s32_f16_m(svint32_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.i32f16(<vscale x 4 x i32> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f16,_m,)(inactive, pg, op);
+}
+
+svint32_t test_svcvt_s32_f32_m(svint32_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f32_m
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.nxv4i32.nxv4f32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f32,_m,)(inactive, pg, op);
+}
+
+svint32_t test_svcvt_s32_f64_m(svint32_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.i32f64(<vscale x 4 x i32> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f64,_m,)(inactive, pg, op);
+}
+
+svint32_t test_svcvt_s32_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.i32f16(<vscale x 4 x i32> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f16,_x,)(pg, op);
+}
+
+svint32_t test_svcvt_s32_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f32_x
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.nxv4i32.nxv4f32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f32,_x,)(pg, op);
+}
+
+svint32_t test_svcvt_s32_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_s32_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzs.i32f64(<vscale x 4 x i32> undef, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s32,_f64,_x,)(pg, op);
+}
+
+svint64_t test_svcvt_s64_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f16_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.i64f16(<vscale x 2 x i64> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f16,_z,)(pg, op);
+}
+
+svint64_t test_svcvt_s64_f32_z(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.i64f32(<vscale x 2 x i64> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f32,_z,)(pg, op);
+}
+
+svint64_t test_svcvt_s64_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f64_z
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.nxv2i64.nxv2f64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f64,_z,)(pg, op);
+}
+
+svint64_t test_svcvt_s64_f16_m(svint64_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.i64f16(<vscale x 2 x i64> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f16,_m,)(inactive, pg, op);
+}
+
+svint64_t test_svcvt_s64_f32_m(svint64_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.i64f32(<vscale x 2 x i64> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f32,_m,)(inactive, pg, op);
+}
+
+svint64_t test_svcvt_s64_f64_m(svint64_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f64_m
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.nxv2i64.nxv2f64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f64,_m,)(inactive, pg, op);
+}
+
+svint64_t test_svcvt_s64_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.i64f16(<vscale x 2 x i64> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f16,_x,)(pg, op);
+}
+
+svint64_t test_svcvt_s64_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.i64f32(<vscale x 2 x i64> undef, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f32,_x,)(pg, op);
+}
+
+svint64_t test_svcvt_s64_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_s64_f64_x
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzs.nxv2i64.nxv2f64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_s64,_f64,_x,)(pg, op);
+}
+
+svuint32_t test_svcvt_u32_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f16_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.i32f16(<vscale x 4 x i32> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f16,_z,)(pg, op);
+}
+
+svuint32_t test_svcvt_u32_f32_z(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f32_z
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.nxv4i32.nxv4f32(<vscale x 4 x i32> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f32,_z,)(pg, op);
+}
+
+svuint32_t test_svcvt_u32_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.i32f64(<vscale x 4 x i32> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f64,_z,)(pg, op);
+}
+
+svuint32_t test_svcvt_u32_f16_m(svuint32_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.i32f16(<vscale x 4 x i32> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f16,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcvt_u32_f32_m(svuint32_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f32_m
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.nxv4i32.nxv4f32(<vscale x 4 x i32> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f32,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcvt_u32_f64_m(svuint32_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.i32f64(<vscale x 4 x i32> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f64,_m,)(inactive, pg, op);
+}
+
+svuint32_t test_svcvt_u32_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.i32f16(<vscale x 4 x i32> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f16,_x,)(pg, op);
+}
+
+svuint32_t test_svcvt_u32_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f32_x
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.nxv4i32.nxv4f32(<vscale x 4 x i32> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f32,_x,)(pg, op);
+}
+
+svuint32_t test_svcvt_u32_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_u32_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.fcvtzu.i32f64(<vscale x 4 x i32> undef, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u32,_f64,_x,)(pg, op);
+}
+
+svuint64_t test_svcvt_u64_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f16_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.i64f16(<vscale x 2 x i64> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f16,_z,)(pg, op);
+}
+
+svuint64_t test_svcvt_u64_f32_z(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.i64f32(<vscale x 2 x i64> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f32,_z,)(pg, op);
+}
+
+svuint64_t test_svcvt_u64_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f64_z
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.nxv2i64.nxv2f64(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f64,_z,)(pg, op);
+}
+
+svuint64_t test_svcvt_u64_f16_m(svuint64_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.i64f16(<vscale x 2 x i64> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f16,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcvt_u64_f32_m(svuint64_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.i64f32(<vscale x 2 x i64> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f32,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcvt_u64_f64_m(svuint64_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f64_m
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.nxv2i64.nxv2f64(<vscale x 2 x i64> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f64,_m,)(inactive, pg, op);
+}
+
+svuint64_t test_svcvt_u64_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.i64f16(<vscale x 2 x i64> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f16,_x,)(pg, op);
+}
+
+svuint64_t test_svcvt_u64_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.i64f32(<vscale x 2 x i64> undef, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f32,_x,)(pg, op);
+}
+
+svuint64_t test_svcvt_u64_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_u64_f64_x
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.fcvtzu.nxv2i64.nxv2f64(<vscale x 2 x i64> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_u64,_f64,_x,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_s32_z(svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_s32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i32(<vscale x 8 x half> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_s32,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_s32_z(svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_s32_z
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.nxv4f32.nxv4i32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_s32,_z,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_s32_z(svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_s32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.f64i32(<vscale x 2 x double> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_s32,_z,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_s32_m(svfloat16_t inactive, svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_s32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i32(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_s32,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvt_f32_s32_m(svfloat32_t inactive, svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_s32_m
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.nxv4f32.nxv4i32(<vscale x 4 x float> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_s32,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvt_f64_s32_m(svfloat64_t inactive, svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_s32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.f64i32(<vscale x 2 x double> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_s32,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvt_f16_s32_x(svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_s32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i32(<vscale x 8 x half> undef, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_s32,_x,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_s32_x(svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_s32_x
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.nxv4f32.nxv4i32(<vscale x 4 x float> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_s32,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_s32_x(svbool_t pg, svint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_s32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.f64i32(<vscale x 2 x double> undef, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_s32,_x,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_s64_z(svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_s64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i64(<vscale x 8 x half> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_s64,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_s64_z(svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_s64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.f32i64(<vscale x 4 x float> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_s64,_z,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_s64_z(svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_s64_z
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.nxv2f64.nxv2i64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_s64,_z,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_s64_m(svfloat16_t inactive, svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_s64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i64(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_s64,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvt_f32_s64_m(svfloat32_t inactive, svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_s64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.f32i64(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_s64,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvt_f64_s64_m(svfloat64_t inactive, svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_s64_m
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.nxv2f64.nxv2i64(<vscale x 2 x double> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_s64,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvt_f16_s64_x(svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_s64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.scvtf.f16i64(<vscale x 8 x half> undef, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_s64,_x,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_s64_x(svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_s64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.scvtf.f32i64(<vscale x 4 x float> undef, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_s64,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_s64_x(svbool_t pg, svint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_s64_x
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.scvtf.nxv2f64.nxv2i64(<vscale x 2 x double> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_s64,_x,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_u32_z(svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_u32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i32(<vscale x 8 x half> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_u32,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_u32_z(svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_u32_z
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.nxv4f32.nxv4i32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_u32,_z,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_u32_z(svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_u32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.f64i32(<vscale x 2 x double> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_u32,_z,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_u32_m(svfloat16_t inactive, svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_u32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i32(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_u32,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvt_f32_u32_m(svfloat32_t inactive, svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_u32_m
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.nxv4f32.nxv4i32(<vscale x 4 x float> %inactive, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_u32,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvt_f64_u32_m(svfloat64_t inactive, svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_u32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.f64i32(<vscale x 2 x double> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_u32,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvt_f16_u32_x(svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_u32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i32(<vscale x 8 x half> undef, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_u32,_x,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_u32_x(svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_u32_x
+  // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.nxv4f32.nxv4i32(<vscale x 4 x float> undef, <vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_u32,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_u32_x(svbool_t pg, svuint32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_u32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.f64i32(<vscale x 2 x double> undef, <vscale x 16 x i1> %pg, <vscale x 4 x i32> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_u32,_x,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_u64_z(svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_u64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i64(<vscale x 8 x half> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_u64,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_u64_z(svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_u64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.f32i64(<vscale x 4 x float> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_u64,_z,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_u64_z(svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_u64_z
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.nxv2f64.nxv2i64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_u64,_z,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_u64_m(svfloat16_t inactive, svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_u64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i64(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_u64,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvt_f32_u64_m(svfloat32_t inactive, svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_u64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.f32i64(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_u64,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvt_f64_u64_m(svfloat64_t inactive, svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_u64_m
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.nxv2f64.nxv2i64(<vscale x 2 x double> %inactive, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_u64,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvt_f16_u64_x(svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_u64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ucvtf.f16i64(<vscale x 8 x half> undef, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_u64,_x,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_u64_x(svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_u64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ucvtf.f32i64(<vscale x 4 x float> undef, <vscale x 16 x i1> %pg, <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_u64,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_u64_x(svbool_t pg, svuint64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_u64_x
+  // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ucvtf.nxv2f64.nxv2i64(<vscale x 2 x double> undef, <vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_u64,_x,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_f16_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f16(<vscale x 4 x float> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_f16,_z,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_f16_z(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_f16_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f16(<vscale x 2 x double> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_f16,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_f16_m(svfloat32_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f16(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_f16,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvt_f64_f16_m(svfloat64_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f16(<vscale x 2 x double> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_f16,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvt_f32_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f16(<vscale x 4 x float> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_f16,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f16(<vscale x 2 x double> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_f16,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_f32_z(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_f32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f32(<vscale x 2 x double> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_f32,_z,)(pg, op);
+}
+
+svfloat64_t test_svcvt_f64_f32_m(svfloat64_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_f32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f32(<vscale x 2 x double> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_f32,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvt_f64_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f64_f32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvt.f64f32(<vscale x 2 x double> undef, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f64,_f32,_x,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_f32_z(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_f32_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f32(<vscale x 8 x half> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_f32,_z,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_f64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f64(<vscale x 8 x half> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_f64,_z,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_f32_m(svfloat16_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_f32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f32(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_f32,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvt_f16_f64_m(svfloat16_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f64(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_f64,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvt_f16_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_f32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f32(<vscale x 8 x half> undef, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_f32,_x,)(pg, op);
+}
+
+svfloat16_t test_svcvt_f16_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f16_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvt.f16f64(<vscale x 8 x half> undef, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f16,_f64,_x,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_f64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f64(<vscale x 4 x float> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_f64,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvt_f32_f64_m(svfloat32_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f64(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_f64,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvt_f32_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvt_f32_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvt.f32f64(<vscale x 4 x float> undef, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  return SVE_ACLE_FUNC(svcvt_f32,_f64,_x,)(pg, op);
+}

clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_cvtlt.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svfloat32_t test_svcvtlt_f32_f16_m(svfloat32_t inactive, svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvtlt_f32_f16_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtlt.f32f16(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtlt_f32_m'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtlt_f32_f16_m'}}
+  return SVE_ACLE_FUNC(svcvtlt_f32,_f16,_m,)(inactive, pg, op);
+}
+
+svfloat64_t test_svcvtlt_f64_f32_m(svfloat64_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvtlt_f64_f32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvtlt.f64f32(<vscale x 2 x double> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtlt_f64_m'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtlt_f64_f32_m'}}
+  return SVE_ACLE_FUNC(svcvtlt_f64,_f32,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvtlt_f32_f16_x(svbool_t pg, svfloat16_t op)
+{
+  // CHECK-LABEL: test_svcvtlt_f32_f16_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtlt.f32f16(<vscale x 4 x float> undef, <vscale x 16 x i1> %pg, <vscale x 8 x half> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtlt_f32_x'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtlt_f32_f16_x'}}
+  return SVE_ACLE_FUNC(svcvtlt_f32,_f16,_x,)(pg, op);
+}
+
+svfloat64_t test_svcvtlt_f64_f32_x(svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvtlt_f64_f32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.fcvtlt.f64f32(<vscale x 2 x double> undef, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 2 x double> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtlt_f64_x'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtlt_f64_f32_x'}}
+  return SVE_ACLE_FUNC(svcvtlt_f64,_f32,_x,)(pg, op);
+}

clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_cvtnt.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svfloat16_t test_svcvtnt_f16_f32_m(svfloat16_t inactive, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvtnt_f16_f32_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvtnt.f16f32(<vscale x 8 x half> %inactive, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtnt_f16_m'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtnt_f16_f32_m'}}
+  return SVE_ACLE_FUNC(svcvtnt_f16,_f32,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvtnt_f32_f64_m(svfloat32_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtnt_f32_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtnt.f32f64(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtnt_f32_m'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtnt_f32_f64_m'}}
+  return SVE_ACLE_FUNC(svcvtnt_f32,_f64,_m,)(inactive, pg, op);
+}
+
+svfloat16_t test_svcvtnt_f16_f32_x(svfloat16_t even, svbool_t pg, svfloat32_t op)
+{
+  // CHECK-LABEL: test_svcvtnt_f16_f32_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.fcvtnt.f16f32(<vscale x 8 x half> %even, <vscale x 16 x i1> %pg, <vscale x 4 x float> %op)
+  // CHECK: ret <vscale x 8 x half> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtnt_f16_x'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtnt_f16_f32_x'}}
+  return SVE_ACLE_FUNC(svcvtnt_f16,_f32,_x,)(even, pg, op);
+}
+
+svfloat32_t test_svcvtnt_f32_f64_x(svfloat32_t even, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtnt_f32_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtnt.f32f64(<vscale x 4 x float> %even, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtnt_f32_x'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtnt_f32_f64_x'}}
+  return SVE_ACLE_FUNC(svcvtnt_f32,_f64,_x,)(even, pg, op);
+}

clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_cvtx.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svfloat32_t test_svcvtx_f32_f64_z(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtx_f32_f64_z
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtx.f32f64(<vscale x 4 x float> zeroinitializer, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtx_f32_z'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtx_f32_f64_z'}}
+  return SVE_ACLE_FUNC(svcvtx_f32,_f64,_z,)(pg, op);
+}
+
+svfloat32_t test_svcvtx_f32_f64_m(svfloat32_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtx_f32_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtx.f32f64(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtx_f32_m'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtx_f32_f64_m'}}
+  return SVE_ACLE_FUNC(svcvtx_f32,_f64,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvtx_f32_f64_x(svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtx_f32_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtx.f32f64(<vscale x 4 x float> undef, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtx_f32_x'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtx_f32_f64_x'}}
+  return SVE_ACLE_FUNC(svcvtx_f32,_f64,_x,)(pg, op);
+}

clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_cvtxnt.c

+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svfloat32_t test_svcvtxnt_f32_f64_m(svfloat32_t inactive, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtxnt_f32_f64_m
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtxnt.f32f64(<vscale x 4 x float> %inactive, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtxnt_f32_m'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtxnt_f32_f64_m'}}
+  return SVE_ACLE_FUNC(svcvtxnt_f32,_f64,_m,)(inactive, pg, op);
+}
+
+svfloat32_t test_svcvtxnt_f32_f64_x(svfloat32_t even, svbool_t pg, svfloat64_t op)
+{
+  // CHECK-LABEL: test_svcvtxnt_f32_f64_x
+  // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.fcvtxnt.f32f64(<vscale x 4 x float> %even, <vscale x 16 x i1> %pg, <vscale x 2 x double> %op)
+  // CHECK: ret <vscale x 4 x float> %[[INTRINSIC]]
+  // overload-warning@+2 {{implicit declaration of function 'svcvtxnt_f32_x'}}
+  // expected-warning@+1 {{implicit declaration of function 'svcvtxnt_f32_f64_x'}}
+  return SVE_ACLE_FUNC(svcvtxnt_f32,_f64,_x,)(even, pg, op);
+}

clang/utils/TableGen/SveEmitter.cpp

case 'm':
     break;
   case 'n':
     Predicate = false;
     Signed = false;
     Float = false;
     ElementBitwidth = Bitwidth = 64;
     NumVectors = 0;
     break;
+  case 'O':
+    Predicate = false;
+    Float = true;
+    ElementBitwidth = 16;
+    break;
+  case 'M':
+    Predicate = false;
+    Float = true;
+    ElementBitwidth = 32;
+    break;
+  case 'N':
+    Predicate = false;
+    Float = true;
+    ElementBitwidth = 64;
+    break;
   case 'S':
     Constant = true;
     Pointer = true;
     ElementBitwidth = Bitwidth = 8;
     NumVectors = 0;
     Signed = true;
     break;
   case 'W':
for (auto &I : Defs) {
 
     // Actually emit the intrinsic declaration.
     I->emitIntrinsic(OS);
   }
 
   if (!InGuard.empty())
     OS << "#endif  //" << InGuard << "\n";
 
+  OS << "#if defined(__ARM_FEATURE_SVE2)\n";
+  OS << "#define svcvtnt_f16_x      svcvtnt_f16_m\n";
+  OS << "#define svcvtnt_f16_f32_x  svcvtnt_f16_f32_m\n";
+  OS << "#define svcvtnt_f32_x      svcvtnt_f32_m\n";
+  OS << "#define svcvtnt_f32_f64_x  svcvtnt_f32_f64_m\n\n";
+
+  OS << "#define svcvtxnt_f32_x     svcvtxnt_f32_m\n";
+  OS << "#define svcvtxnt_f32_f64_x svcvtxnt_f32_f64_m\n\n";
+
+  OS << "#endif /*__ARM_FEATURE_SVE2 */\n\n";
+
   OS << "#ifdef __cplusplus\n";
   OS << "} // extern \"C\"\n";
   OS << "#endif\n\n";
   OS << "#endif /*__ARM_FEATURE_SVE */\n\n";
   OS << "#endif /* __ARM_SVE_H */\n";
 }
 
 void SVEEmitter::createBuiltins(raw_ostream &OS) {