[Clang][AArch64] Use generic extract/insert vector for svget/svset/svcreate tuples

[CarolineConcatto]

Summary

This patch replaces svget, svset and svcreate aarch64 intrinsics for tuple
types with the generic llvm-ir intrinsics extract/insert vector

[RKSimon]

Are you intending to AutoUpgrade the existing intrinsics in IR?

[nlopes]

clang/lib/CodeGen/CGBuiltin.cpp

9115

Please use PoisonValue here

[sdesmalen]

clang/lib/CodeGen/CGBuiltin.cpp

9078

nit: s/Emmit/Emit/

How about naming these:

• EmitSVETupleGet
• EmitSVETupleSet
• EmitSVETupleCreate

9085

nit: s/unsigned int/unsigned/

9092

nit: Can this function be merged with the function above, you could pass bool IsInsert?

[CarolineConcatto]

• Have EmitSVETupleSetOrGet for isTupleSet and isTupleGet
• Address review comments

[CarolineConcatto]

Are you intending to AutoUpgrade the existing intrinsics in IR?

@RKSimon
AFAIK we want to remove them altogether.
It is not expected that external projects depend on these intrinsics.
It is a legacy intrinsic that was introduced before we had vector.extract and vector.insert

clang/lib/CodeGen/CGBuiltin.cpp

9092

I did what you suggested. Don't know if it is much better.

[RKSimon]

In D131547#3712509, @CarolineConcatto wrote:

Are you intending to AutoUpgrade the existing intrinsics in IR?

@RKSimon
AFAIK we want to remove them altogether.
It is not expected that external projects depend on these intrinsics.
It is a legacy intrinsic that was introduced before we had vector.extract and vector.insert

OK - not sure what aarch64 policy is, but x86 we tend to continue support for old intrinsics that could still be in IR through AutoUpgrade.cpp which mostly converts them to generics (very similar to CGBuiltin) - technically we plan to remove the oldest at some point but we've been very loathe to actually do it.

[dmgreen]

I think we usually try to do the same, if the intrinsics have been in released compilers. There is an example in https://reviews.llvm.org/D98487#change-tOTTgECYYAO5, hopefully these would be equally simple.

[sdesmalen]

In D131547#3723083, @dmgreen wrote:

I think we usually try to do the same, if the intrinsics have been in released compilers. There is an example in https://reviews.llvm.org/D98487#change-tOTTgECYYAO5, hopefully these would be equally simple.

We don't really have the intention in keeping compatibility for intrinsics like these, since the vector.extract/insert intrinsics are the proper way to insert/extract vectors and they have been around for quite some time now. Also these intrinsics are quite specific to the initial SVE ACLE implementation when we first upstreamed it, so it's not very likely that anyone else is using them.

clang/lib/CodeGen/CGBuiltin.cpp

9085–9097

I think you could simplify this further to:

auto *SingleVecTy = cast<llvm::ScalableVectorType>(
    TypeFlags.isTupleSet() ? Ops[2]->getType() : Ty);

unsigned I = cast<ConstantInt>(Ops[1])->getSExtValue();
Value *Idx =
    ConstantInt::get(CGM.Int64Ty, I * SingleVecTy->getMinNumElements());
  
if (TypeFlags.isTupleSet())
  return Builder.CreateExtractVector(Ty, Ops[0], Ops[2], Idx);
else
  return Builder.CreateInsertVector(Ty, Ops[0], Idx);

9106

Casting to VectorType here isn't actually necessary, because both llvm::PoisonValue::get() and Builder.CreateInsertVector() take a Type*.

9107

nit: If you cast this to llvm::ScalableVectorType you can call SrcTy->getMinNumElements() directly.

[dmgreen]

In D131547#3728842, @sdesmalen wrote:

In D131547#3723083, @dmgreen wrote:

I think we usually try to do the same, if the intrinsics have been in released compilers. There is an example in https://reviews.llvm.org/D98487#change-tOTTgECYYAO5, hopefully these would be equally simple.

We don't really have the intention in keeping compatibility for intrinsics like these, since the vector.extract/insert intrinsics are the proper way to insert/extract vectors and they have been around for quite some time now. Also these intrinsics are quite specific to the initial SVE ACLE implementation when we first upstreamed it, so it's not very likely that anyone else is using them.

I'm not sure what you mean by "keeping compatibility", they just need to be replaced in the autoupdate code. If there was a release of llvm which emitted the intrinsics, then the default assumption is that someone might have bitcode that uses them. Luckily it is really simple to add the autoupgrade code in most cases.

[sdesmalen]

In D131547#3728867, @dmgreen wrote:

In D131547#3728842, @sdesmalen wrote:

In D131547#3723083, @dmgreen wrote:

I think we usually try to do the same, if the intrinsics have been in released compilers. There is an example in https://reviews.llvm.org/D98487#change-tOTTgECYYAO5, hopefully these would be equally simple.

We don't really have the intention in keeping compatibility for intrinsics like these, since the vector.extract/insert intrinsics are the proper way to insert/extract vectors and they have been around for quite some time now. Also these intrinsics are quite specific to the initial SVE ACLE implementation when we first upstreamed it, so it's not very likely that anyone else is using them.

I'm not sure what you mean by "keeping compatibility", they just need to be replaced in the autoupdate code. If there was a release of llvm which emitted the intrinsics, then the default assumption is that someone might have bitcode that uses them.

What I meant with 'keeping compatibility' is exactly what you described; bitcode/IR that uses the old intrinsics remaining to work with newer versions of LLVM.

My point was that these intrinsics have not been in LLVM for that long and have a very limited scope, so are unlikely to have accrued much legacy. I don't really expect a practical use-case where people have legacy SVE ACLE bitcode that they need to compile with a newer version of LLVM. So unless someone explicitly requests the compatibility, we'd rather remove them in favour of adding code that will never be used.

Is there a formal requirement that LLVM must remain backward compatible with older LLVM IR (beyond the target-independent parts)?

[CarolineConcatto]

• Address nit in EmitSVETupleSetOrGet

[dmgreen]

Is there a formal requirement that LLVM must remain backward compatible with older LLVM IR (beyond the target-independent parts)?

We have always done it in the past, and I don't see a good reason to change. This change is essentially for llvm 16, so we are talking about any bitcode between when SVE was added and that release. It is hard to tell how people will use bitcode up to that point and if they will expect it to continue working going forward. I think it's simpler to just add the upgrade code, than to try and argue that it is unneeded. But the upgrade code is really needed in D131548 (and D131687) where the old intrinsics are being removed.

[sdesmalen]

Thanks for addressing the comments @CarolineConcatto!

In D131547#3731310, @dmgreen wrote:

Is there a formal requirement that LLVM must remain backward compatible with older LLVM IR (beyond the target-independent parts)?

We have always done it in the past, and I don't see a good reason to change. This change is essentially for llvm 16, so we are talking about any bitcode between when SVE was added and that release. It is hard to tell how people will use bitcode up to that point and if they will expect it to continue working going forward. I think it's simpler to just add the upgrade code, than to try and argue that it is unneeded. But the upgrade code is really needed in D131548 (and D131687) where the old intrinsics are being removed.

It seems that the LLVM Developer Policy provides better guidance than that.

Newer releases can ignore features from older releases, but they cannot miscompile them

Removing the intrinsics but not auto-upgrading them would mean that older IR would miscompile (the call to the intrinsic would become an actual function call). This suggests there is no freedom of choice here and we must use AutoUpgrade.

clang/lib/CodeGen/CGBuiltin.cpp

9099

nit: redundant newline, please remove.

[CarolineConcatto]

• nit

[sdesmalen]

LGTM, thanks @CarolineConcatto. Please land D132137 before landing this one to avoid regressions on combines that previously worked on svget/svset.

clang/lib/CodeGen/CGBuiltin.cpp

9382

nit: If you cast this to ScalableVectorType, you can use VTy->getMinNumElements() below on line 9385

clang/include/clang/Basic/arm_sve.td

 
 ////////////////////////////////////////////////////////////////////////////////
 // Vector creation
 def SVUNDEF_1 : SInst<"svundef_{d}",  "dv", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef]>;
 def SVUNDEF_2 : SInst<"svundef2_{d}", "2v", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef]>;
 def SVUNDEF_3 : SInst<"svundef3_{d}", "3v", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef]>;
 def SVUNDEF_4 : SInst<"svundef4_{d}", "4v", "csilUcUsUiUlhfd", MergeNone, "", [IsUndef]>;
 
-def SVCREATE_2 : SInst<"svcreate2[_{d}]", "2dd",   "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_create2", [IsTupleCreate]>;
-def SVCREATE_3 : SInst<"svcreate3[_{d}]", "3ddd",  "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_create3", [IsTupleCreate]>;
-def SVCREATE_4 : SInst<"svcreate4[_{d}]", "4dddd", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_create4", [IsTupleCreate]>;
+def SVCREATE_2 : SInst<"svcreate2[_{d}]", "2dd",   "csilUcUsUiUlhfd", MergeNone, "", [IsTupleCreate]>;
+def SVCREATE_3 : SInst<"svcreate3[_{d}]", "3ddd",  "csilUcUsUiUlhfd", MergeNone, "", [IsTupleCreate]>;
+def SVCREATE_4 : SInst<"svcreate4[_{d}]", "4dddd", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleCreate]>;
 
 let ArchGuard = "defined(__ARM_FEATURE_SVE_BF16)" in {
 def SVUNDEF_1_BF16 : SInst<"svundef_{d}",  "dv", "b", MergeNone, "", [IsUndef]>;
 def SVUNDEF_2_BF16 : SInst<"svundef2_{d}", "2v", "b", MergeNone, "", [IsUndef]>;
 def SVUNDEF_3_BF16 : SInst<"svundef3_{d}", "3v", "b", MergeNone, "", [IsUndef]>;
 def SVUNDEF_4_BF16 : SInst<"svundef4_{d}", "4v", "b", MergeNone, "", [IsUndef]>;
 
-def SVCREATE_2_BF16 : SInst<"svcreate2[_{d}]", "2dd",   "b", MergeNone, "aarch64_sve_tuple_create2", [IsTupleCreate]>;
-def SVCREATE_3_BF16 : SInst<"svcreate3[_{d}]", "3ddd",  "b", MergeNone, "aarch64_sve_tuple_create3", [IsTupleCreate]>;
-def SVCREATE_4_BF16 : SInst<"svcreate4[_{d}]", "4dddd", "b", MergeNone, "aarch64_sve_tuple_create4", [IsTupleCreate]>;
+def SVCREATE_2_BF16 : SInst<"svcreate2[_{d}]", "2dd",   "b", MergeNone, "", [IsTupleCreate]>;
+def SVCREATE_3_BF16 : SInst<"svcreate3[_{d}]", "3ddd",  "b", MergeNone, "", [IsTupleCreate]>;
+def SVCREATE_4_BF16 : SInst<"svcreate4[_{d}]", "4dddd", "b", MergeNone, "", [IsTupleCreate]>;
 }
 
 ////////////////////////////////////////////////////////////////////////////////
 // Vector insertion and extraction
-def SVGET_2 : SInst<"svget2[_{d}]", "d2i", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_get", [IsTupleGet], [ImmCheck<1, ImmCheck0_1>]>;
-def SVGET_3 : SInst<"svget3[_{d}]", "d3i", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_get", [IsTupleGet], [ImmCheck<1, ImmCheck0_2>]>;
-def SVGET_4 : SInst<"svget4[_{d}]", "d4i", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_get", [IsTupleGet], [ImmCheck<1, ImmCheck0_3>]>;
-
-def SVSET_2 : SInst<"svset2[_{d}]", "22id", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_set", [IsTupleSet], [ImmCheck<1, ImmCheck0_1>]>;
-def SVSET_3 : SInst<"svset3[_{d}]", "33id", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_set", [IsTupleSet], [ImmCheck<1, ImmCheck0_2>]>;
-def SVSET_4 : SInst<"svset4[_{d}]", "44id", "csilUcUsUiUlhfd", MergeNone, "aarch64_sve_tuple_set", [IsTupleSet], [ImmCheck<1, ImmCheck0_3>]>;
-
-let ArchGuard = "defined(__ARM_FEATURE_SVE_BF16)" in {
-def SVGET_2_BF16 : SInst<"svget2[_{d}]", "d2i", "b", MergeNone, "aarch64_sve_tuple_get", [IsTupleGet], [ImmCheck<1, ImmCheck0_1>]>;
-def SVGET_3_BF16 : SInst<"svget3[_{d}]", "d3i", "b", MergeNone, "aarch64_sve_tuple_get", [IsTupleGet], [ImmCheck<1, ImmCheck0_2>]>;
-def SVGET_4_BF16 : SInst<"svget4[_{d}]", "d4i", "b", MergeNone, "aarch64_sve_tuple_get", [IsTupleGet], [ImmCheck<1, ImmCheck0_3>]>;
-
-def SVSET_2_BF16 : SInst<"svset2[_{d}]", "22id", "b", MergeNone, "aarch64_sve_tuple_set", [IsTupleSet], [ImmCheck<1, ImmCheck0_1>]>;
-def SVSET_3_BF16 : SInst<"svset3[_{d}]", "33id", "b", MergeNone, "aarch64_sve_tuple_set", [IsTupleSet], [ImmCheck<1, ImmCheck0_2>]>;
-def SVSET_4_BF16 : SInst<"svset4[_{d}]", "44id", "b", MergeNone, "aarch64_sve_tuple_set", [IsTupleSet], [ImmCheck<1, ImmCheck0_3>]>;
+def SVGET_2 : SInst<"svget2[_{d}]", "d2i", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleGet], [ImmCheck<1, ImmCheck0_1>]>;
+def SVGET_3 : SInst<"svget3[_{d}]", "d3i", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleGet], [ImmCheck<1, ImmCheck0_2>]>;
+def SVGET_4 : SInst<"svget4[_{d}]", "d4i", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleGet], [ImmCheck<1, ImmCheck0_3>]>;
+
+def SVSET_2 : SInst<"svset2[_{d}]", "22id", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleSet], [ImmCheck<1, ImmCheck0_1>]>;
+def SVSET_3 : SInst<"svset3[_{d}]", "33id", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleSet], [ImmCheck<1, ImmCheck0_2>]>;
+def SVSET_4 : SInst<"svset4[_{d}]", "44id", "csilUcUsUiUlhfd", MergeNone, "", [IsTupleSet], [ImmCheck<1, ImmCheck0_3>]>;
+
+let ArchGuard = "defined(__ARM_FEATURE_SVE_BF16)" in {
+def SVGET_2_BF16 : SInst<"svget2[_{d}]", "d2i", "b", MergeNone, "", [IsTupleGet], [ImmCheck<1, ImmCheck0_1>]>;
+def SVGET_3_BF16 : SInst<"svget3[_{d}]", "d3i", "b", MergeNone, "", [IsTupleGet], [ImmCheck<1, ImmCheck0_2>]>;
+def SVGET_4_BF16 : SInst<"svget4[_{d}]", "d4i", "b", MergeNone, "", [IsTupleGet], [ImmCheck<1, ImmCheck0_3>]>;
+
+def SVSET_2_BF16 : SInst<"svset2[_{d}]", "22id", "b", MergeNone, "", [IsTupleSet], [ImmCheck<1, ImmCheck0_1>]>;
+def SVSET_3_BF16 : SInst<"svset3[_{d}]", "33id", "b", MergeNone, "", [IsTupleSet], [ImmCheck<1, ImmCheck0_2>]>;
+def SVSET_4_BF16 : SInst<"svset4[_{d}]", "44id", "b", MergeNone, "", [IsTupleSet], [ImmCheck<1, ImmCheck0_3>]>;
 }
 
 ////////////////////////////////////////////////////////////////////////////////
 // SVE2 WhileGE/GT
 let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
 def SVWHILEGE_S32 : SInst<"svwhilege_{d}[_{1}]", "Pkk", "PcPsPiPl",     MergeNone, "aarch64_sve_whilege", [IsOverloadWhile]>;
 def SVWHILEGE_S64 : SInst<"svwhilege_{d}[_{1}]", "Pll", "PcPsPiPl",     MergeNone, "aarch64_sve_whilege", [IsOverloadWhile]>;
 def SVWHILEGT_S32 : SInst<"svwhilegt_{d}[_{1}]", "Pkk", "PcPsPiPl",     MergeNone, "aarch64_sve_whilegt", [IsOverloadWhile]>;

clang/lib/CodeGen/CGBuiltin.cpp

 //===---- CGBuiltin.cpp - Emit LLVM Code for builtins ---------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // This contains code to emit Builtin calls as LLVM code.
case Intrinsic::aarch64_sve_st3:
     N = 3;
     break;
   case Intrinsic::aarch64_sve_st4:
     N = 4;
     break;
   default:
     llvm_unreachable("unknown intrinsic!");
   }
-  auto TupleTy =
-      llvm::VectorType::get(VTy->getElementType(), VTy->getElementCount() * N);
 
   Value *Predicate = EmitSVEPredicateCast(Ops[0], VTy);
   Value *BasePtr = Builder.CreateBitCast(Ops[1], VecPtrTy);
   Value *Offset = Ops.size() > 3 ? Ops[2] : Builder.getInt32(0);
   Value *Val = Ops.back();
   BasePtr = Builder.CreateGEP(VTy, BasePtr, Offset);
   BasePtr = Builder.CreateBitCast(BasePtr, EltPtrTy);
 
   // The llvm.aarch64.sve.st2/3/4 intrinsics take legal part vectors, so we
   // need to break up the tuple vector.
   SmallVector<llvm::Value*, 5> Operands;
-  Function *FExtr =
-      CGM.getIntrinsic(Intrinsic::aarch64_sve_tuple_get, {VTy, TupleTy});
-  for (unsigned I = 0; I < N; ++I)
-    Operands.push_back(Builder.CreateCall(FExtr, {Val, Builder.getInt32(I)}));
+  unsigned MinElts = VTy->getElementCount().getKnownMinValue();
+  for (unsigned I = 0; I < N; ++I) {
+    Value *Idx = ConstantInt::get(CGM.Int64Ty, I * MinElts);
+    Operands.push_back(Builder.CreateExtractVector(VTy, Val, Idx));
+  }
   Operands.append({Predicate, BasePtr});
 
   Function *F = CGM.getIntrinsic(IntID, { VTy });
   return Builder.CreateCall(F, Operands);
 }
 
 // SVE2's svpmullb and svpmullt builtins are similar to the svpmullb_pair and
 // svpmullt_pair intrinsics, with the exception that their results are bitcast
CodeGenFunction::getSVEOverloadTypes(const 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() || TypeFlags.isTupleSet())
+  if (TypeFlags.isOverloadCvt())
     return {Ops[0]->getType(), Ops.back()->getType()};
 
-  if (TypeFlags.isTupleCreate() || TypeFlags.isTupleGet())
-    return {ResultType, Ops[0]->getType()};
-
   assert(TypeFlags.isOverloadDefault() && "Unexpected value for overloads");
   return {DefaultType};
 }
 
+Value *CodeGenFunction::EmitSVETupleSetOrGet(const SVETypeFlags &TypeFlags,
+                                             llvm::Type *Ty,
+                                             ArrayRef<Value *> Ops) {
+  assert((TypeFlags.isTupleSet() || TypeFlags.isTupleGet()) &&
+         "Expects TypleFlag isTupleSet or TypeFlags.isTupleSet()");
+
+  unsigned I = cast<ConstantInt>(Ops[1])->getSExtValue();
+  auto *SingleVecTy = dyn_cast<llvm::ScalableVectorType>(
+                      TypeFlags.isTupleSet() ? Ops[2]->getType() : Ty);
+  Value *Idx = ConstantInt::get(CGM.Int64Ty,
+                                I * SingleVecTy->getMinNumElements());
+
+  if (TypeFlags.isTupleSet())
+    return Builder.CreateInsertVector(Ty, Ops[0], Ops[2], Idx);
+  return Builder.CreateExtractVector(Ty, Ops[0], Idx);
+}
+
+Value *CodeGenFunction::EmitSVETupleCreate(const SVETypeFlags &TypeFlags,
+                                             llvm::Type *Ty,
+                                             ArrayRef<Value *> Ops) {
+  assert(TypeFlags.isTupleCreate() && "Expects TypleFlag isTupleCreate");
+
+  auto *SrcTy = dyn_cast<llvm::ScalableVectorType>(Ops[0]->getType());
+  unsigned MinElts = SrcTy->getMinNumElements();
+  Value *Call = llvm::PoisonValue::get(Ty);
+  for (unsigned I = 0; I < Ops.size(); I++) {
+    Value *Idx = ConstantInt::get(CGM.Int64Ty, I * MinElts);
+    Call = Builder.CreateInsertVector(Ty, Call, Ops[I], Idx);
+  }
+
+  return Call;
+}
+
 Value *CodeGenFunction::EmitAArch64SVEBuiltinExpr(unsigned BuiltinID,
                                                   const CallExpr *E) {
   // Find out if any arguments are required to be integer constant expressions.
   unsigned ICEArguments = 0;
   ASTContext::GetBuiltinTypeError Error;
   getContext().GetBuiltinType(BuiltinID, Error, &ICEArguments);
   assert(Error == ASTContext::GE_None && "Should not codegen an error");
 
   llvm::Type *Ty = ConvertType(E->getType());
   if (BuiltinID >= SVE::BI__builtin_sve_reinterpret_s8_s8 &&
       BuiltinID <= SVE::BI__builtin_sve_reinterpret_f64_f64) {
     Value *Val = EmitScalarExpr(E->getArg(0));
     return EmitSVEReinterpret(Val, Ty);
Value *CodeGenFunction::EmitAArch64SVEBuiltinExpr(unsigned BuiltinID,
   else if (TypeFlags.isPrefetch())
     return EmitSVEPrefetchLoad(TypeFlags, Ops, Builtin->LLVMIntrinsic);
   else if (TypeFlags.isGatherPrefetch())
     return EmitSVEGatherPrefetch(TypeFlags, Ops, Builtin->LLVMIntrinsic);
   else if (TypeFlags.isStructLoad())
     return EmitSVEStructLoad(TypeFlags, Ops, Builtin->LLVMIntrinsic);
   else if (TypeFlags.isStructStore())
     return EmitSVEStructStore(TypeFlags, Ops, Builtin->LLVMIntrinsic);
+  else if (TypeFlags.isTupleSet() || TypeFlags.isTupleGet())
+        return EmitSVETupleSetOrGet(TypeFlags, Ty, Ops);
+  else if (TypeFlags.isTupleCreate())
+        return EmitSVETupleCreate(TypeFlags, Ty, Ops);
   else if (TypeFlags.isUndef())
     return UndefValue::get(Ty);
   else if (Builtin->LLVMIntrinsic != 0) {
     if (TypeFlags.getMergeType() == SVETypeFlags::MergeZeroExp)
       InsertExplicitZeroOperand(Builder, Ty, Ops);
 
     if (TypeFlags.getMergeType() == SVETypeFlags::MergeAnyExp)
       InsertExplicitUndefOperand(Builder, Ty, Ops);
Value *CodeGenFunction::EmitAArch64SVEBuiltinExpr(unsigned BuiltinID,
   case SVE::BI__builtin_sve_svtbl2_s32:
   case SVE::BI__builtin_sve_svtbl2_u64:
   case SVE::BI__builtin_sve_svtbl2_s64:
   case SVE::BI__builtin_sve_svtbl2_f16:
   case SVE::BI__builtin_sve_svtbl2_bf16:
   case SVE::BI__builtin_sve_svtbl2_f32:
   case SVE::BI__builtin_sve_svtbl2_f64: {
     SVETypeFlags TF(Builtin->TypeModifier);
     auto VTy = cast<llvm::VectorType>(getSVEType(TF));
-    auto TupleTy = llvm::VectorType::getDoubleElementsVectorType(VTy);
-    Function *FExtr =
-        CGM.getIntrinsic(Intrinsic::aarch64_sve_tuple_get, {VTy, TupleTy});
-    Value *V0 = Builder.CreateCall(FExtr, {Ops[0], Builder.getInt32(0)});
-    Value *V1 = Builder.CreateCall(FExtr, {Ops[0], Builder.getInt32(1)});
+    Value *V0 = Builder.CreateExtractVector(VTy, Ops[0],
+                                            ConstantInt::get(CGM.Int64Ty, 0));
+    unsigned MinElts = VTy->getElementCount().getKnownMinValue();
+    Value *V1 = Builder.CreateExtractVector(
+        VTy, Ops[0], ConstantInt::get(CGM.Int64Ty, MinElts));
     Function *F = CGM.getIntrinsic(Intrinsic::aarch64_sve_tbl2, VTy);
     return Builder.CreateCall(F, {V0, V1, Ops[1]});
   }
 
   case SVE::BI__builtin_sve_svset_neonq_s8:
   case SVE::BI__builtin_sve_svset_neonq_s16:
   case SVE::BI__builtin_sve_svset_neonq_s32:
   case SVE::BI__builtin_sve_svset_neonq_s64:

clang/lib/CodeGen/CodeGenFunction.h

public:
   llvm::Type *SVEBuiltinMemEltTy(const SVETypeFlags &TypeFlags);
 
   SmallVector<llvm::Type *, 2>
   getSVEOverloadTypes(const SVETypeFlags &TypeFlags, llvm::Type *ReturnType,
                       ArrayRef<llvm::Value *> Ops);
   llvm::Type *getEltType(const SVETypeFlags &TypeFlags);
   llvm::ScalableVectorType *getSVEType(const SVETypeFlags &TypeFlags);
   llvm::ScalableVectorType *getSVEPredType(const SVETypeFlags &TypeFlags);
+  llvm::Value *EmitSVETupleSetOrGet(const SVETypeFlags &TypeFlags,
+                                    llvm::Type *ReturnType,
+                                    ArrayRef<llvm::Value *> Ops);
+  llvm::Value *EmitSVETupleCreate(const SVETypeFlags &TypeFlags,
+                                  llvm::Type *ReturnType,
+                                  ArrayRef<llvm::Value *> Ops);
   llvm::Value *EmitSVEAllTruePred(const SVETypeFlags &TypeFlags);
   llvm::Value *EmitSVEDupX(llvm::Value *Scalar);
   llvm::Value *EmitSVEDupX(llvm::Value *Scalar, llvm::Type *Ty);
   llvm::Value *EmitSVEReinterpret(llvm::Value *Val, llvm::Type *Ty);
   llvm::Value *EmitSVEPMull(const SVETypeFlags &TypeFlags,
                             llvm::SmallVectorImpl<llvm::Value *> &Ops,
                             unsigned BuiltinID);
   llvm::Value *EmitSVEMovl(const SVETypeFlags &TypeFlags,

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create2-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svcreate2_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.aarch64.sve.tuple.create2.nxv16bf16.nxv8bf16(<vscale x 8 x bfloat> [[X0:%.*]], <vscale x 8 x bfloat> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svcreate2_bf16u14__SVBFloat16_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.aarch64.sve.tuple.create2.nxv16bf16.nxv8bf16(<vscale x 8 x bfloat> [[X0:%.*]], <vscale x 8 x bfloat> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP1]]
 //
 svbfloat16x2_t test_svcreate2_bf16(svbfloat16_t x0, svbfloat16_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_bf16,,)(x0, x1);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create2.c

 // 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
 
 // CHECK-LABEL: @test_svcreate2_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.create2.nxv32i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z17test_svcreate2_s8u10__SVInt8_tu10__SVInt8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.create2.nxv32i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP1]]
 //
 svint8x2_t test_svcreate2_s8(svint8_t x0, svint8_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_s8,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.create2.nxv16i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_s16u11__SVInt16_tu11__SVInt16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.create2.nxv16i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
 //
 svint16x2_t test_svcreate2_s16(svint16_t x0, svint16_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_s16,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.create2.nxv8i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_s32u11__SVInt32_tu11__SVInt32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.create2.nxv8i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
 //
 svint32x2_t test_svcreate2_s32(svint32_t x0, svint32_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_s32,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.create2.nxv4i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_s64u11__SVInt64_tu11__SVInt64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.create2.nxv4i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
 //
 svint64x2_t test_svcreate2_s64(svint64_t x0, svint64_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_s64,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.create2.nxv32i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z17test_svcreate2_u8u11__SVUint8_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.create2.nxv32i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP1]]
 //
 svuint8x2_t test_svcreate2_u8(svuint8_t x0, svuint8_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_u8,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.create2.nxv16i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_u16u12__SVUint16_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.create2.nxv16i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
 //
 svuint16x2_t test_svcreate2_u16(svuint16_t x0, svuint16_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_u16,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.create2.nxv8i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_u32u12__SVUint32_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.create2.nxv8i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
 //
 svuint32x2_t test_svcreate2_u32(svuint32_t x0, svuint32_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_u32,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.create2.nxv4i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_u64u12__SVUint64_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.create2.nxv4i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
 //
 svuint64x2_t test_svcreate2_u64(svuint64_t x0, svuint64_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_u64,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.aarch64.sve.tuple.create2.nxv16f16.nxv8f16(<vscale x 8 x half> [[X0:%.*]], <vscale x 8 x half> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x half> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x half> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_f16u13__SVFloat16_tu13__SVFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.aarch64.sve.tuple.create2.nxv16f16.nxv8f16(<vscale x 8 x half> [[X0:%.*]], <vscale x 8 x half> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP1]]
 //
 svfloat16x2_t test_svcreate2_f16(svfloat16_t x0, svfloat16_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_f16,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.aarch64.sve.tuple.create2.nxv8f32.nxv4f32(<vscale x 4 x float> [[X0:%.*]], <vscale x 4 x float> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x float> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x float> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_f32u13__SVFloat32_tu13__SVFloat32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.aarch64.sve.tuple.create2.nxv8f32.nxv4f32(<vscale x 4 x float> [[X0:%.*]], <vscale x 4 x float> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP1]]
 //
 svfloat32x2_t test_svcreate2_f32(svfloat32_t x0, svfloat32_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_f32,,)(x0, x1);
 }
 
 // CHECK-LABEL: @test_svcreate2_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.aarch64.sve.tuple.create2.nxv4f64.nxv2f64(<vscale x 2 x double> [[X0:%.*]], <vscale x 2 x double> [[X1:%.*]])
-// CHECK-NEXT:    ret <vscale x 4 x double> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x double> [[TMP1]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate2_f64u13__SVFloat64_tu13__SVFloat64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.aarch64.sve.tuple.create2.nxv4f64.nxv2f64(<vscale x 2 x double> [[X0:%.*]], <vscale x 2 x double> [[X1:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP1]]
 //
 svfloat64x2_t test_svcreate2_f64(svfloat64_t x0, svfloat64_t x1)
 {
   return SVE_ACLE_FUNC(svcreate2,_f64,,)(x0, x1);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create3-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svcreate3_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.create3.nxv24bf16.nxv8bf16(<vscale x 8 x bfloat> [[X0:%.*]], <vscale x 8 x bfloat> [[X1:%.*]], <vscale x 8 x bfloat> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svcreate3_bf16u14__SVBFloat16_tu14__SVBFloat16_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.create3.nxv24bf16.nxv8bf16(<vscale x 8 x bfloat> [[X0:%.*]], <vscale x 8 x bfloat> [[X1:%.*]], <vscale x 8 x bfloat> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP2]]
 //
 svbfloat16x3_t test_svcreate3_bf16(svbfloat16_t x0, svbfloat16_t x1, svbfloat16_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_bf16,,)(x0, x1, x2);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create3.c

 // 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
 
 // CHECK-LABEL: @test_svcreate3_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.create3.nxv48i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z17test_svcreate3_s8u10__SVInt8_tu10__SVInt8_tu10__SVInt8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.create3.nxv48i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
 //
 svint8x3_t test_svcreate3_s8(svint8_t x0, svint8_t x1, svint8_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_s8,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.create3.nxv24i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_s16u11__SVInt16_tu11__SVInt16_tu11__SVInt16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.create3.nxv24i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP2]]
 //
 svint16x3_t test_svcreate3_s16(svint16_t x0, svint16_t x1, svint16_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_s16,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.create3.nxv12i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_s32u11__SVInt32_tu11__SVInt32_tu11__SVInt32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.create3.nxv12i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP2]]
 //
 svint32x3_t test_svcreate3_s32(svint32_t x0, svint32_t x1, svint32_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_s32,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.create3.nxv6i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_s64u11__SVInt64_tu11__SVInt64_tu11__SVInt64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.create3.nxv6i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP2]]
 //
 svint64x3_t test_svcreate3_s64(svint64_t x0, svint64_t x1, svint64_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_s64,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.create3.nxv48i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z17test_svcreate3_u8u11__SVUint8_tu11__SVUint8_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.create3.nxv48i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
 //
 svuint8x3_t test_svcreate3_u8(svuint8_t x0, svuint8_t x1, svuint8_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_u8,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.create3.nxv24i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_u16u12__SVUint16_tu12__SVUint16_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.create3.nxv24i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP2]]
 //
 svuint16x3_t test_svcreate3_u16(svuint16_t x0, svuint16_t x1, svuint16_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_u16,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.create3.nxv12i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_u32u12__SVUint32_tu12__SVUint32_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.create3.nxv12i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP2]]
 //
 svuint32x3_t test_svcreate3_u32(svuint32_t x0, svuint32_t x1, svuint32_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_u32,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.create3.nxv6i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_u64u12__SVUint64_tu12__SVUint64_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.create3.nxv6i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP2]]
 //
 svuint64x3_t test_svcreate3_u64(svuint64_t x0, svuint64_t x1, svuint64_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_u64,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.aarch64.sve.tuple.create3.nxv24f16.nxv8f16(<vscale x 8 x half> [[X0:%.*]], <vscale x 8 x half> [[X1:%.*]], <vscale x 8 x half> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x half> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x half> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_f16u13__SVFloat16_tu13__SVFloat16_tu13__SVFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.aarch64.sve.tuple.create3.nxv24f16.nxv8f16(<vscale x 8 x half> [[X0:%.*]], <vscale x 8 x half> [[X1:%.*]], <vscale x 8 x half> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP2]]
 //
 svfloat16x3_t test_svcreate3_f16(svfloat16_t x0, svfloat16_t x1, svfloat16_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_f16,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.aarch64.sve.tuple.create3.nxv12f32.nxv4f32(<vscale x 4 x float> [[X0:%.*]], <vscale x 4 x float> [[X1:%.*]], <vscale x 4 x float> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 12 x float> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x float> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_f32u13__SVFloat32_tu13__SVFloat32_tu13__SVFloat32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.aarch64.sve.tuple.create3.nxv12f32.nxv4f32(<vscale x 4 x float> [[X0:%.*]], <vscale x 4 x float> [[X1:%.*]], <vscale x 4 x float> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP2]]
 //
 svfloat32x3_t test_svcreate3_f32(svfloat32_t x0, svfloat32_t x1, svfloat32_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_f32,,)(x0, x1, x2);
 }
 
 // CHECK-LABEL: @test_svcreate3_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.aarch64.sve.tuple.create3.nxv6f64.nxv2f64(<vscale x 2 x double> [[X0:%.*]], <vscale x 2 x double> [[X1:%.*]], <vscale x 2 x double> [[X2:%.*]])
-// CHECK-NEXT:    ret <vscale x 6 x double> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x double> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate3_f64u13__SVFloat64_tu13__SVFloat64_tu13__SVFloat64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.aarch64.sve.tuple.create3.nxv6f64.nxv2f64(<vscale x 2 x double> [[X0:%.*]], <vscale x 2 x double> [[X1:%.*]], <vscale x 2 x double> [[X2:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP2]]
 //
 svfloat64x3_t test_svcreate3_f64(svfloat64_t x0, svfloat64_t x1, svfloat64_t x2)
 {
   return SVE_ACLE_FUNC(svcreate3,_f64,,)(x0, x1, x2);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create4-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svcreate4_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.create4.nxv32bf16.nxv8bf16(<vscale x 8 x bfloat> [[X0:%.*]], <vscale x 8 x bfloat> [[X1:%.*]], <vscale x 8 x bfloat> [[X2:%.*]], <vscale x 8 x bfloat> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[X4:%.*]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svcreate4_bf16u14__SVBFloat16_tu14__SVBFloat16_tu14__SVBFloat16_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.create4.nxv32bf16.nxv8bf16(<vscale x 8 x bfloat> [[X0:%.*]], <vscale x 8 x bfloat> [[X1:%.*]], <vscale x 8 x bfloat> [[X2:%.*]], <vscale x 8 x bfloat> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[X4:%.*]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP3]]
 //
 svbfloat16x4_t test_svcreate4_bf16(svbfloat16_t x0, svbfloat16_t x1, svbfloat16_t x2, svbfloat16_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_bf16,,)(x0, x1, x2, x4);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_create4.c

 // 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
 
 // CHECK-LABEL: @test_svcreate4_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.create4.nxv64i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]], <vscale x 16 x i8> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
+// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z17test_svcreate4_s8u10__SVInt8_tu10__SVInt8_tu10__SVInt8_tu10__SVInt8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.create4.nxv64i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]], <vscale x 16 x i8> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
+// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP3]]
 //
 svint8x4_t test_svcreate4_s8(svint8_t x0, svint8_t x1, svint8_t x2, svint8_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_s8,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.create4.nxv32i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]], <vscale x 8 x i16> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_s16u11__SVInt16_tu11__SVInt16_tu11__SVInt16_tu11__SVInt16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.create4.nxv32i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]], <vscale x 8 x i16> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
 //
 svint16x4_t test_svcreate4_s16(svint16_t x0, svint16_t x1, svint16_t x2, svint16_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_s16,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.create4.nxv16i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]], <vscale x 4 x i32> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_s32u11__SVInt32_tu11__SVInt32_tu11__SVInt32_tu11__SVInt32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.create4.nxv16i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]], <vscale x 4 x i32> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
 //
 svint32x4_t test_svcreate4_s32(svint32_t x0, svint32_t x1, svint32_t x2, svint32_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_s32,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.create4.nxv8i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]], <vscale x 2 x i64> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_s64u11__SVInt64_tu11__SVInt64_tu11__SVInt64_tu11__SVInt64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.create4.nxv8i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]], <vscale x 2 x i64> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
 //
 svint64x4_t test_svcreate4_s64(svint64_t x0, svint64_t x1, svint64_t x2, svint64_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_s64,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.create4.nxv64i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]], <vscale x 16 x i8> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
+// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z17test_svcreate4_u8u11__SVUint8_tu11__SVUint8_tu11__SVUint8_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.create4.nxv64i8.nxv16i8(<vscale x 16 x i8> [[X0:%.*]], <vscale x 16 x i8> [[X1:%.*]], <vscale x 16 x i8> [[X2:%.*]], <vscale x 16 x i8> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP0]], <vscale x 16 x i8> [[X1:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP1]], <vscale x 16 x i8> [[X2:%.*]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[X4:%.*]], i64 48)
+// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP3]]
 //
 svuint8x4_t test_svcreate4_u8(svuint8_t x0, svuint8_t x1, svuint8_t x2, svuint8_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_u8,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.create4.nxv32i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]], <vscale x 8 x i16> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_u16u12__SVUint16_tu12__SVUint16_tu12__SVUint16_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.create4.nxv32i16.nxv8i16(<vscale x 8 x i16> [[X0:%.*]], <vscale x 8 x i16> [[X1:%.*]], <vscale x 8 x i16> [[X2:%.*]], <vscale x 8 x i16> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP0]], <vscale x 8 x i16> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP1]], <vscale x 8 x i16> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP2]], <vscale x 8 x i16> [[X4:%.*]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
 //
 svuint16x4_t test_svcreate4_u16(svuint16_t x0, svuint16_t x1, svuint16_t x2, svuint16_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_u16,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.create4.nxv16i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]], <vscale x 4 x i32> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_u32u12__SVUint32_tu12__SVUint32_tu12__SVUint32_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.create4.nxv16i32.nxv4i32(<vscale x 4 x i32> [[X0:%.*]], <vscale x 4 x i32> [[X1:%.*]], <vscale x 4 x i32> [[X2:%.*]], <vscale x 4 x i32> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP0]], <vscale x 4 x i32> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP1]], <vscale x 4 x i32> [[X2:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP2]], <vscale x 4 x i32> [[X4:%.*]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
 //
 svuint32x4_t test_svcreate4_u32(svuint32_t x0, svuint32_t x1, svuint32_t x2, svuint32_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_u32,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.create4.nxv8i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]], <vscale x 2 x i64> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_u64u12__SVUint64_tu12__SVUint64_tu12__SVUint64_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.create4.nxv8i64.nxv2i64(<vscale x 2 x i64> [[X0:%.*]], <vscale x 2 x i64> [[X1:%.*]], <vscale x 2 x i64> [[X2:%.*]], <vscale x 2 x i64> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP0]], <vscale x 2 x i64> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP1]], <vscale x 2 x i64> [[X2:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP2]], <vscale x 2 x i64> [[X4:%.*]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
 //
 svuint64x4_t test_svcreate4_u64(svuint64_t x0, svuint64_t x1, svuint64_t x2, svuint64_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_u64,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.aarch64.sve.tuple.create4.nxv32f16.nxv8f16(<vscale x 8 x half> [[X0:%.*]], <vscale x 8 x half> [[X1:%.*]], <vscale x 8 x half> [[X2:%.*]], <vscale x 8 x half> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x half> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[X4:%.*]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x half> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_f16u13__SVFloat16_tu13__SVFloat16_tu13__SVFloat16_tu13__SVFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.aarch64.sve.tuple.create4.nxv32f16.nxv8f16(<vscale x 8 x half> [[X0:%.*]], <vscale x 8 x half> [[X1:%.*]], <vscale x 8 x half> [[X2:%.*]], <vscale x 8 x half> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP0]], <vscale x 8 x half> [[X1:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP1]], <vscale x 8 x half> [[X2:%.*]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP2]], <vscale x 8 x half> [[X4:%.*]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP3]]
 //
 svfloat16x4_t test_svcreate4_f16(svfloat16_t x0, svfloat16_t x1, svfloat16_t x2, svfloat16_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_f16,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.aarch64.sve.tuple.create4.nxv16f32.nxv4f32(<vscale x 4 x float> [[X0:%.*]], <vscale x 4 x float> [[X1:%.*]], <vscale x 4 x float> [[X2:%.*]], <vscale x 4 x float> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x float> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[X4:%.*]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x float> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_f32u13__SVFloat32_tu13__SVFloat32_tu13__SVFloat32_tu13__SVFloat32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.aarch64.sve.tuple.create4.nxv16f32.nxv4f32(<vscale x 4 x float> [[X0:%.*]], <vscale x 4 x float> [[X1:%.*]], <vscale x 4 x float> [[X2:%.*]], <vscale x 4 x float> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP0]], <vscale x 4 x float> [[X1:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP1]], <vscale x 4 x float> [[X2:%.*]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP2]], <vscale x 4 x float> [[X4:%.*]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP3]]
 //
 svfloat32x4_t test_svcreate4_f32(svfloat32_t x0, svfloat32_t x1, svfloat32_t x2, svfloat32_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_f32,,)(x0, x1, x2, x4);
 }
 
 // CHECK-LABEL: @test_svcreate4_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.aarch64.sve.tuple.create4.nxv8f64.nxv2f64(<vscale x 2 x double> [[X0:%.*]], <vscale x 2 x double> [[X1:%.*]], <vscale x 2 x double> [[X2:%.*]], <vscale x 2 x double> [[X4:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x double> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[X4:%.*]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x double> [[TMP3]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svcreate4_f64u13__SVFloat64_tu13__SVFloat64_tu13__SVFloat64_tu13__SVFloat64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.aarch64.sve.tuple.create4.nxv8f64.nxv2f64(<vscale x 2 x double> [[X0:%.*]], <vscale x 2 x double> [[X1:%.*]], <vscale x 2 x double> [[X2:%.*]], <vscale x 2 x double> [[X4:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[X0:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP0]], <vscale x 2 x double> [[X1:%.*]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP1]], <vscale x 2 x double> [[X2:%.*]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP2]], <vscale x 2 x double> [[X4:%.*]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP3]]
 //
 svfloat64x4_t test_svcreate4_f64(svfloat64_t x0, svfloat64_t x1, svfloat64_t x2, svfloat64_t x4)
 {
   return SVE_ACLE_FUNC(svcreate4,_f64,,)(x0, x1, x2, x4);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get2-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svget2_bf16_0(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget2_bf16_014svbfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget2_bf16_0(svbfloat16x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_bf16,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget2_bf16_1(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget2_bf16_114svbfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget2_bf16_1(svbfloat16x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_bf16,,)(tuple, 1);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get2.c

 // 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
 
 // CHECK-LABEL: @test_svget2_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svget2_s810svint8x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 svint8_t test_svget2_s8(svint8x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_s8,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget2_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_s1611svint16x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 svint16_t test_svget2_s16(svint16x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_s16,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget2_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_s3211svint32x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 svint32_t test_svget2_s32(svint32x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_s32,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget2_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_s6411svint64x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 svint64_t test_svget2_s64(svint64x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_s64,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget2_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svget2_u811svuint8x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 svuint8_t test_svget2_u8(svuint8x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_u8,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget2_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_u1612svuint16x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 svuint16_t test_svget2_u16(svuint16x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_u16,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget2_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_u3212svuint32x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 svuint32_t test_svget2_u32(svuint32x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_u32,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget2_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_u6412svuint64x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[TUPLE:%.*]], i64 2)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 svuint64_t test_svget2_u64(svuint64x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_u64,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget2_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_f1613svfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
 //
 svfloat16_t test_svget2_f16(svfloat16x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_f16,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget2_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[TUPLE:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_f3213svfloat32x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[TUPLE:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
 //
 svfloat32_t test_svget2_f32(svfloat32x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_f32,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget2_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 2 x double> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget2_f6413svfloat64x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x double> [[TMP0]]
 //
 svfloat64_t test_svget2_f64(svfloat64x2_t tuple)
 {
   return SVE_ACLE_FUNC(svget2,_f64,,)(tuple, 0);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get3-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svget3_bf16_0(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget3_bf16_014svbfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget3_bf16_0(svbfloat16x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_bf16,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget3_bf16_1(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget3_bf16_114svbfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget3_bf16_1(svbfloat16x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_bf16,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget3_bf16_2(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget3_bf16_214svbfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget3_bf16_2(svbfloat16x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_bf16,,)(tuple, 2);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get3.c

 // 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
 
 // CHECK-LABEL: @test_svget3_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svget3_s810svint8x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 svint8_t test_svget3_s8(svint8x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_s8,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget3_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_s1611svint16x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 svint16_t test_svget3_s16(svint16x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_s16,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget3_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_s3211svint32x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 svint32_t test_svget3_s32(svint32x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_s32,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget3_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_s6411svint64x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 svint64_t test_svget3_s64(svint64x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_s64,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget3_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 32)
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svget3_u811svuint8x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[TUPLE:%.*]], i64 32)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 svuint8_t test_svget3_u8(svuint8x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_u8,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget3_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_u1612svuint16x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 svuint16_t test_svget3_u16(svuint16x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_u16,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget3_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_u3212svuint32x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 svuint32_t test_svget3_u32(svuint32x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_u32,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget3_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_u6412svuint64x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[TUPLE:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 svuint64_t test_svget3_u64(svuint64x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_u64,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget3_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_f1613svfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
 //
 svfloat16_t test_svget3_f16(svfloat16x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_f16,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget3_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_f3213svfloat32x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
 //
 svfloat32_t test_svget3_f32(svfloat32x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_f32,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget3_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[TUPLE:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 2 x double> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget3_f6413svfloat64x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[TUPLE:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x double> [[TMP0]]
 //
 svfloat64_t test_svget3_f64(svfloat64x3_t tuple)
 {
   return SVE_ACLE_FUNC(svget3,_f64,,)(tuple, 2);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get4-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svget4_bf16_0(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget4_bf16_014svbfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget4_bf16_0(svbfloat16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_bf16,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget4_bf16_1(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget4_bf16_114svbfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget4_bf16_1(svbfloat16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_bf16,,)(tuple, 1);
 }
 
 // CHECK-LABEL: @test_svget4_bf16_2(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget4_bf16_214svbfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget4_bf16_2(svbfloat16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_bf16,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget4_bf16_3(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 3)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 24)
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svget4_bf16_314svbfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i64 24)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
 //
 svbfloat16_t test_svget4_bf16_3(svbfloat16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_bf16,,)(tuple, 3);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_get4.c

 #else
 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
 #endif
 
 // NOTE: For these tests clang converts the struct parameter into
 // several parameters, one for each member of the original struct.
 // CHECK-LABEL: @test_svget4_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svget4_s810svint8x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 svint8_t test_svget4_s8(svint8x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_s8,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget4_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_s1611svint16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 svint16_t test_svget4_s16(svint16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_s16,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget4_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_s3211svint32x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 svint32_t test_svget4_s32(svint32x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_s32,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget4_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 3)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_s6411svint64x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 svint64_t test_svget4_s64(svint64x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_s64,,)(tuple, 3);
 }
 
 // CHECK-LABEL: @test_svget4_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 32)
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svget4_u811svuint8x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[TUPLE:%.*]], i64 32)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
 //
 svuint8_t test_svget4_u8(svuint8x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_u8,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget4_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 3)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 24)
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_u1612svuint16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[TUPLE:%.*]], i64 24)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
 //
 svuint16_t test_svget4_u16(svuint16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_u16,,)(tuple, 3);
 }
 
 // CHECK-LABEL: @test_svget4_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_u3212svuint32x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
 //
 svuint32_t test_svget4_u32(svuint32x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_u32,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget4_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 3)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_u6412svuint64x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[TUPLE:%.*]], i64 6)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
 //
 svuint64_t test_svget4_u64(svuint64x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_u64,,)(tuple, 3);
 }
 
 // CHECK-LABEL: @test_svget4_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[TUPLE:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_f1613svfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[TUPLE:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
 //
 svfloat16_t test_svget4_f16(svfloat16x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_f16,,)(tuple, 2);
 }
 
 // CHECK-LABEL: @test_svget4_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[TUPLE:%.*]], i32 0)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[TUPLE:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_f3213svfloat32x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[TUPLE:%.*]], i32 0)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[TUPLE:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
 //
 svfloat32_t test_svget4_f32(svfloat32x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_f32,,)(tuple, 0);
 }
 
 // CHECK-LABEL: @test_svget4_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[TUPLE:%.*]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[TUPLE:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 2 x double> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svget4_f6413svfloat64x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[TUPLE:%.*]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[TUPLE:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 2 x double> [[TMP0]]
 //
 svfloat64_t test_svget4_f64(svfloat64x4_t tuple)
 {
   return SVE_ACLE_FUNC(svget4,_f64,,)(tuple, 2);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set2-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svset2_bf16_0(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.aarch64.sve.tuple.set.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 0, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset2_bf16_014svbfloat16x2_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.aarch64.sve.tuple.set.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 0, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
 //
 svbfloat16x2_t test_svset2_bf16_0(svbfloat16x2_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset2,_bf16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset2_bf16_1(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.aarch64.sve.tuple.set.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 1, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset2_bf16_114svbfloat16x2_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.aarch64.sve.tuple.set.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], i32 1, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP0]]
 //
 svbfloat16x2_t test_svset2_bf16_1(svbfloat16x2_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset2,_bf16,,)(tuple, 1, x);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set2.c

 // 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
 
 // CHECK-LABEL: @test_svset2_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.set.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svset2_s810svint8x2_tu10__SVInt8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.set.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
 //
 svint8x2_t test_svset2_s8(svint8x2_t tuple, svint8_t x)
 {
   return SVE_ACLE_FUNC(svset2,_s8,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset2_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.set.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 0, <vscale x 8 x i16> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_s1611svint16x2_tu11__SVInt16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.set.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 0, <vscale x 8 x i16> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
 //
 svint16x2_t test_svset2_s16(svint16x2_t tuple, svint16_t x)
 {
   return SVE_ACLE_FUNC(svset2,_s16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset2_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.set.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_s3211svint32x2_tu11__SVInt32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.set.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
 //
 svint32x2_t test_svset2_s32(svint32x2_t tuple, svint32_t x)
 {
   return SVE_ACLE_FUNC(svset2,_s32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset2_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.set.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 0, <vscale x 2 x i64> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_s6411svint64x2_tu11__SVInt64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.set.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 0, <vscale x 2 x i64> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
 //
 svint64x2_t test_svset2_s64(svint64x2_t tuple, svint64_t x)
 {
   return SVE_ACLE_FUNC(svset2,_s64,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset2_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.set.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svset2_u811svuint8x2_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.tuple.set.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
 //
 svuint8x2_t test_svset2_u8(svuint8x2_t tuple, svuint8_t x)
 {
   return SVE_ACLE_FUNC(svset2,_u8,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset2_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.set.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 0, <vscale x 8 x i16> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_u1612svuint16x2_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.aarch64.sve.tuple.set.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], i32 0, <vscale x 8 x i16> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP0]]
 //
 svuint16x2_t test_svset2_u16(svuint16x2_t tuple, svuint16_t x)
 {
   return SVE_ACLE_FUNC(svset2,_u16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset2_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.set.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_u3212svuint32x2_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.aarch64.sve.tuple.set.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP0]]
 //
 svuint32x2_t test_svset2_u32(svuint32x2_t tuple, svuint32_t x)
 {
   return SVE_ACLE_FUNC(svset2,_u32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset2_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.set.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 0, <vscale x 2 x i64> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_u6412svuint64x2_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.aarch64.sve.tuple.set.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], i32 0, <vscale x 2 x i64> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP0]]
 //
 svuint64x2_t test_svset2_u64(svuint64x2_t tuple, svuint64_t x)
 {
   return SVE_ACLE_FUNC(svset2,_u64,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset2_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.aarch64.sve.tuple.set.nxv16f16.nxv8f16(<vscale x 16 x half> [[TUPLE:%.*]], i32 1, <vscale x 8 x half> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 16 x half> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_f1613svfloat16x2_tu13__SVFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.aarch64.sve.tuple.set.nxv16f16.nxv8f16(<vscale x 16 x half> [[TUPLE:%.*]], i32 1, <vscale x 8 x half> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP0]]
 //
 svfloat16x2_t test_svset2_f16(svfloat16x2_t tuple, svfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset2,_f16,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset2_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.aarch64.sve.tuple.set.nxv8f32.nxv4f32(<vscale x 8 x float> [[TUPLE:%.*]], i32 0, <vscale x 4 x float> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 8 x float> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_f3213svfloat32x2_tu13__SVFloat32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.aarch64.sve.tuple.set.nxv8f32.nxv4f32(<vscale x 8 x float> [[TUPLE:%.*]], i32 0, <vscale x 4 x float> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP0]]
 //
 svfloat32x2_t test_svset2_f32(svfloat32x2_t tuple, svfloat32_t x)
 {
   return SVE_ACLE_FUNC(svset2,_f32,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset2_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.aarch64.sve.tuple.set.nxv4f64.nxv2f64(<vscale x 4 x double> [[TUPLE:%.*]], i32 1, <vscale x 2 x double> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 2)
 // CHECK-NEXT:    ret <vscale x 4 x double> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset2_f6413svfloat64x2_tu13__SVFloat64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.aarch64.sve.tuple.set.nxv4f64.nxv2f64(<vscale x 4 x double> [[TUPLE:%.*]], i32 1, <vscale x 2 x double> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 2)
 // CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP0]]
 //
 svfloat64x2_t test_svset2_f64(svfloat64x2_t tuple, svfloat64_t x)
 {
   return SVE_ACLE_FUNC(svset2,_f64,,)(tuple, 1, x);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set3-bfloat.c

 #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
 
 
 // CHECK-LABEL: @test_svset3_bf16_0(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.set.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 0, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset3_bf16_014svbfloat16x3_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.set.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 0, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
 //
 svbfloat16x3_t test_svset3_bf16_0(svbfloat16x3_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset3,_bf16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset3_bf16_1(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.set.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 1, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset3_bf16_114svbfloat16x3_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.set.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 1, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
 //
 svbfloat16x3_t test_svset3_bf16_1(svbfloat16x3_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset3,_bf16,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset3_bf16_2(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.set.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 2, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset3_bf16_214svbfloat16x3_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.aarch64.sve.tuple.set.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], i32 2, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP0]]
 //
 svbfloat16x3_t test_svset3_bf16_2(svbfloat16x3_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset3,_bf16,,)(tuple, 2, x);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set3.c

 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
 #endif
 
 
 // NOTE: For these tests clang converts the struct parameter into
 // several parameters, one for each member of the original struct.
 // CHECK-LABEL: @test_svset3_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.set.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svset3_s810svint8x3_tu10__SVInt8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.set.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
 //
 svint8x3_t test_svset3_s8(svint8x3_t tuple, svint8_t x)
 {
   return SVE_ACLE_FUNC(svset3,_s8,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset3_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.set.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 2, <vscale x 8 x i16> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_s1611svint16x3_tu11__SVInt16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.set.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 2, <vscale x 8 x i16> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
 //
 svint16x3_t test_svset3_s16(svint16x3_t tuple, svint16_t x)
 {
   return SVE_ACLE_FUNC(svset3,_s16,,)(tuple, 2, x);
 }
 
 // CHECK-LABEL: @test_svset3_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.set.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 0, <vscale x 4 x i32> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_s3211svint32x3_tu11__SVInt32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.set.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 0, <vscale x 4 x i32> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
 //
 svint32x3_t test_svset3_s32(svint32x3_t tuple, svint32_t x)
 {
   return SVE_ACLE_FUNC(svset3,_s32,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset3_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.set.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 1, <vscale x 2 x i64> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
 // CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_s6411svint64x3_tu11__SVInt64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.set.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 1, <vscale x 2 x i64> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
 // CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
 //
 svint64x3_t test_svset3_s64(svint64x3_t tuple, svint64_t x)
 {
   return SVE_ACLE_FUNC(svset3,_s64,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset3_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.set.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 2, <vscale x 16 x i8> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 32)
 // CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svset3_u811svuint8x3_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.tuple.set.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], i32 2, <vscale x 16 x i8> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 32)
 // CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
 //
 svuint8x3_t test_svset3_u8(svuint8x3_t tuple, svuint8_t x)
 {
   return SVE_ACLE_FUNC(svset3,_u8,,)(tuple, 2, x);
 }
 
 // CHECK-LABEL: @test_svset3_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.set.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 0, <vscale x 8 x i16> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_u1612svuint16x3_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.aarch64.sve.tuple.set.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], i32 0, <vscale x 8 x i16> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP0]]
 //
 svuint16x3_t test_svset3_u16(svuint16x3_t tuple, svuint16_t x)
 {
   return SVE_ACLE_FUNC(svset3,_u16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset3_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.set.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_u3212svuint32x3_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.aarch64.sve.tuple.set.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP0]]
 //
 svuint32x3_t test_svset3_u32(svuint32x3_t tuple, svuint32_t x)
 {
   return SVE_ACLE_FUNC(svset3,_u32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset3_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.set.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 2, <vscale x 2 x i64> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_u6412svuint64x3_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.aarch64.sve.tuple.set.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], i32 2, <vscale x 2 x i64> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP0]]
 //
 svuint64x3_t test_svset3_u64(svuint64x3_t tuple, svuint64_t x)
 {
   return SVE_ACLE_FUNC(svset3,_u64,,)(tuple, 2, x);
 }
 
 // CHECK-LABEL: @test_svset3_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.aarch64.sve.tuple.set.nxv24f16.nxv8f16(<vscale x 24 x half> [[TUPLE:%.*]], i32 0, <vscale x 8 x half> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 24 x half> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_f1613svfloat16x3_tu13__SVFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.aarch64.sve.tuple.set.nxv24f16.nxv8f16(<vscale x 24 x half> [[TUPLE:%.*]], i32 0, <vscale x 8 x half> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP0]]
 //
 svfloat16x3_t test_svset3_f16(svfloat16x3_t tuple, svfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset3,_f16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset3_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.aarch64.sve.tuple.set.nxv12f32.nxv4f32(<vscale x 12 x float> [[TUPLE:%.*]], i32 1, <vscale x 4 x float> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 12 x float> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_f3213svfloat32x3_tu13__SVFloat32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.aarch64.sve.tuple.set.nxv12f32.nxv4f32(<vscale x 12 x float> [[TUPLE:%.*]], i32 1, <vscale x 4 x float> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP0]]
 //
 svfloat32x3_t test_svset3_f32(svfloat32x3_t tuple, svfloat32_t x)
 {
   return SVE_ACLE_FUNC(svset3,_f32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset3_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.aarch64.sve.tuple.set.nxv6f64.nxv2f64(<vscale x 6 x double> [[TUPLE:%.*]], i32 2, <vscale x 2 x double> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 6 x double> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset3_f6413svfloat64x3_tu13__SVFloat64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.aarch64.sve.tuple.set.nxv6f64.nxv2f64(<vscale x 6 x double> [[TUPLE:%.*]], i32 2, <vscale x 2 x double> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP0]]
 //
 svfloat64x3_t test_svset3_f64(svfloat64x3_t tuple, svfloat64_t x)
 {
   return SVE_ACLE_FUNC(svset3,_f64,,)(tuple, 2, x);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set4-bfloat.c

 #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
 
 
 // CHECK-LABEL: @test_svset4_bf16_0(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 0, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
 // CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset4_bf16_014svbfloat16x4_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 0, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 0)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 svbfloat16x4_t test_svset4_bf16_0(svbfloat16x4_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_bf16,,)(tuple, 0, x);
 }
 
 // CHECK-LABEL: @test_svset4_bf16_1(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 1, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset4_bf16_114svbfloat16x4_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 1, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 svbfloat16x4_t test_svset4_bf16_1(svbfloat16x4_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_bf16,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_bf16_2(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 2, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset4_bf16_214svbfloat16x4_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 2, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 svbfloat16x4_t test_svset4_bf16_2(svbfloat16x4_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_bf16,,)(tuple, 2, x);
 }
 
 // CHECK-LABEL: @test_svset4_bf16_3(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 3, <vscale x 8 x bfloat> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 24)
 // CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svset4_bf16_314svbfloat16x4_tu14__SVBFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.aarch64.sve.tuple.set.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], i32 3, <vscale x 8 x bfloat> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TUPLE:%.*]], <vscale x 8 x bfloat> [[X:%.*]], i64 24)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP0]]
 //
 svbfloat16x4_t test_svset4_bf16_3(svbfloat16x4_t tuple, svbfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_bf16,,)(tuple, 3, x);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_set4.c

 #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
 
 
 // CHECK-LABEL: @test_svset4_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.set.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svset4_s810svint8x4_tu10__SVInt8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.set.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 1, <vscale x 16 x i8> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 16)
 // CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
 //
 svint8x4_t test_svset4_s8(svint8x4_t tuple, svint8_t x)
 {
   return SVE_ACLE_FUNC(svset4,_s8,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.set.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 3, <vscale x 8 x i16> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 24)
 // CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_s1611svint16x4_tu11__SVInt16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.set.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 3, <vscale x 8 x i16> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 24)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
 //
 svint16x4_t test_svset4_s16(svint16x4_t tuple, svint16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_s16,,)(tuple, 3, x);
 }
 
 // CHECK-LABEL: @test_svset4_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.set.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_s3211svint32x4_tu11__SVInt32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.set.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
 //
 svint32x4_t test_svset4_s32(svint32x4_t tuple, svint32_t x)
 {
   return SVE_ACLE_FUNC(svset4,_s32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.set.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 1, <vscale x 2 x i64> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
 // CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_s6411svint64x4_tu11__SVInt64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.set.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 1, <vscale x 2 x i64> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 2)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
 //
 svint64x4_t test_svset4_s64(svint64x4_t tuple, svint64_t x)
 {
   return SVE_ACLE_FUNC(svset4,_s64,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.set.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 3, <vscale x 16 x i8> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 48)
 // CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svset4_u811svuint8x4_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.tuple.set.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], i32 3, <vscale x 16 x i8> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TUPLE:%.*]], <vscale x 16 x i8> [[X:%.*]], i64 48)
 // CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
 //
 svuint8x4_t test_svset4_u8(svuint8x4_t tuple, svuint8_t x)
 {
   return SVE_ACLE_FUNC(svset4,_u8,,)(tuple, 3, x);
 }
 
 // CHECK-LABEL: @test_svset4_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.set.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 1, <vscale x 8 x i16> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_u1612svuint16x4_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.aarch64.sve.tuple.set.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], i32 1, <vscale x 8 x i16> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TUPLE:%.*]], <vscale x 8 x i16> [[X:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP0]]
 //
 svuint16x4_t test_svset4_u16(svuint16x4_t tuple, svuint16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_u16,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.set.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_u3212svuint32x4_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.aarch64.sve.tuple.set.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], i32 1, <vscale x 4 x i32> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TUPLE:%.*]], <vscale x 4 x i32> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP0]]
 //
 svuint32x4_t test_svset4_u32(svuint32x4_t tuple, svuint32_t x)
 {
   return SVE_ACLE_FUNC(svset4,_u32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.set.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 3, <vscale x 2 x i64> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 6)
 // CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_u6412svuint64x4_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.aarch64.sve.tuple.set.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], i32 3, <vscale x 2 x i64> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TUPLE:%.*]], <vscale x 2 x i64> [[X:%.*]], i64 6)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP0]]
 //
 svuint64x4_t test_svset4_u64(svuint64x4_t tuple, svuint64_t x)
 {
   return SVE_ACLE_FUNC(svset4,_u64,,)(tuple, 3, x);
 }
 
 // CHECK-LABEL: @test_svset4_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.aarch64.sve.tuple.set.nxv32f16.nxv8f16(<vscale x 32 x half> [[TUPLE:%.*]], i32 1, <vscale x 8 x half> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
 // CHECK-NEXT:    ret <vscale x 32 x half> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_f1613svfloat16x4_tu13__SVFloat16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.aarch64.sve.tuple.set.nxv32f16.nxv8f16(<vscale x 32 x half> [[TUPLE:%.*]], i32 1, <vscale x 8 x half> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TUPLE:%.*]], <vscale x 8 x half> [[X:%.*]], i64 8)
 // CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP0]]
 //
 svfloat16x4_t test_svset4_f16(svfloat16x4_t tuple, svfloat16_t x)
 {
   return SVE_ACLE_FUNC(svset4,_f16,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.aarch64.sve.tuple.set.nxv16f32.nxv4f32(<vscale x 16 x float> [[TUPLE:%.*]], i32 1, <vscale x 4 x float> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
 // CHECK-NEXT:    ret <vscale x 16 x float> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_f3213svfloat32x4_tu13__SVFloat32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.aarch64.sve.tuple.set.nxv16f32.nxv4f32(<vscale x 16 x float> [[TUPLE:%.*]], i32 1, <vscale x 4 x float> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TUPLE:%.*]], <vscale x 4 x float> [[X:%.*]], i64 4)
 // CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP0]]
 //
 svfloat32x4_t test_svset4_f32(svfloat32x4_t tuple, svfloat32_t x)
 {
   return SVE_ACLE_FUNC(svset4,_f32,,)(tuple, 1, x);
 }
 
 // CHECK-LABEL: @test_svset4_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.aarch64.sve.tuple.set.nxv8f64.nxv2f64(<vscale x 8 x double> [[TUPLE:%.*]], i32 3, <vscale x 2 x double> [[X:%.*]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 6)
 // CHECK-NEXT:    ret <vscale x 8 x double> [[TMP0]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svset4_f6413svfloat64x4_tu13__SVFloat64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.aarch64.sve.tuple.set.nxv8f64.nxv2f64(<vscale x 8 x double> [[TUPLE:%.*]], i32 3, <vscale x 2 x double> [[X:%.*]])
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TUPLE:%.*]], <vscale x 2 x double> [[X:%.*]], i64 6)
 // CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP0]]
 //
 svfloat64x4_t test_svset4_f64(svfloat64x4_t tuple, svfloat64_t x)
 {
   return SVE_ACLE_FUNC(svset4,_f64,,)(tuple, 3, x);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st2-bfloat.c

 // 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
 // CHECK-LABEL: @test_svst2_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z15test_svst2_bf16u10__SVBool_tPu6__bf1614svbfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_bf16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast bfloat* [[BASE:%.*]] to <vscale x 8 x bfloat>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z20test_svst2_vnum_bf16u10__SVBool_tPu6__bf16l14svbfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast bfloat* [[BASE:%.*]] to <vscale x 8 x bfloat>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_bf16(svbool_t pg, bfloat16_t *base, int64_t vnum, svbfloat16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_bf16,,)(pg, base, vnum, data);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st2.c

 // 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
 
 // CHECK-LABEL: @test_svst2_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z13test_svst2_s8u10__SVBool_tPa10svint8x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_s8(svbool_t pg, int8_t *base, svint8x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_s8,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_s16u10__SVBool_tPs11svint16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_s16(svbool_t pg, int16_t *base, svint16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_s16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_s32u10__SVBool_tPi11svint32x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_s32(svbool_t pg, int32_t *base, svint32x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_s32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_s64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_s64u10__SVBool_tPl11svint64x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_s64(svbool_t pg, int64_t *base, svint64x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_s64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z13test_svst2_u8u10__SVBool_tPh11svuint8x2_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_u8(svbool_t pg, uint8_t *base, svuint8x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_u8,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_u16u10__SVBool_tPt12svuint16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_u16(svbool_t pg, uint16_t *base, svuint16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_u16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_u32u10__SVBool_tPj12svuint32x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_u32(svbool_t pg, uint32_t *base, svuint32x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_u32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_u64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_u64u10__SVBool_tPm12svuint64x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_u64(svbool_t pg, uint64_t *base, svuint64x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_u64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_f16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_f16u10__SVBool_tPDh13svfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_f16(svbool_t pg, float16_t *base, svfloat16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_f16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_f32u10__SVBool_tPf13svfloat32x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_f32(svbool_t pg, float32_t *base, svfloat32x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_f32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst2_f64u10__SVBool_tPd13svfloat64x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_f64(svbool_t pg, float64_t *base, svfloat64x2_t data)
 {
   return SVE_ACLE_FUNC(svst2,_f64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_s8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z18test_svst2_vnum_s8u10__SVBool_tPal10svint8x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_s8,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_s16u10__SVBool_tPsl11svint16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_s16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_s32u10__SVBool_tPil11svint32x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_s32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_s64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_s64u10__SVBool_tPll11svint64x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_s64,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_u8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z18test_svst2_vnum_u8u10__SVBool_tPhl11svuint8x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_u8,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_u16u10__SVBool_tPtl12svuint16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_u16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_u32u10__SVBool_tPjl12svuint32x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_u32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_u64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_u64u10__SVBool_tPml12svuint64x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_u64,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_f16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast half* [[BASE:%.*]] to <vscale x 8 x half>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_f16u10__SVBool_tPDhl13svfloat16x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast half* [[BASE:%.*]] to <vscale x 8 x half>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_f16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[BASE:%.*]] to <vscale x 4 x float>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_f32u10__SVBool_tPfl13svfloat32x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[BASE:%.*]] to <vscale x 4 x float>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_f32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst2_vnum_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[BASE:%.*]] to <vscale x 2 x double>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst2_vnum_f64u10__SVBool_tPdl13svfloat64x2_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[BASE:%.*]] to <vscale x 2 x double>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st2.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst2_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x2_t data)
 {
   return SVE_ACLE_FUNC(svst2_vnum,_f64,,)(pg, base, vnum, data);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st3-bfloat.c

 #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
 
 // CHECK-LABEL: @test_svst3_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z15test_svst3_bf16u10__SVBool_tPu6__bf1614svbfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_bf16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast bfloat* [[BASE:%.*]] to <vscale x 8 x bfloat>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z20test_svst3_vnum_bf16u10__SVBool_tPu6__bf16l14svbfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast bfloat* [[BASE:%.*]] to <vscale x 8 x bfloat>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv24bf16(<vscale x 24 x bfloat> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_bf16(svbool_t pg, bfloat16_t *base, int64_t vnum, svbfloat16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_bf16,,)(pg, base, vnum, data);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st3.c

 // 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
 
 // CHECK-LABEL: @test_svst3_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z13test_svst3_s8u10__SVBool_tPa10svint8x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_s8(svbool_t pg, int8_t *base, svint8x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_s8,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_s16u10__SVBool_tPs11svint16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_s16(svbool_t pg, int16_t *base, svint16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_s16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_s32u10__SVBool_tPi11svint32x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_s32(svbool_t pg, int32_t *base, svint32x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_s32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_s64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_s64u10__SVBool_tPl11svint64x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_s64(svbool_t pg, int64_t *base, svint64x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_s64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z13test_svst3_u8u10__SVBool_tPh11svuint8x3_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_u8(svbool_t pg, uint8_t *base, svuint8x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_u8,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_u16u10__SVBool_tPt12svuint16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_u16(svbool_t pg, uint16_t *base, svuint16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_u16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_u32u10__SVBool_tPj12svuint32x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_u32(svbool_t pg, uint32_t *base, svuint32x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_u32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_u64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_u64u10__SVBool_tPm12svuint64x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_u64(svbool_t pg, uint64_t *base, svuint64x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_u64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_f16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_f16u10__SVBool_tPDh13svfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_f16(svbool_t pg, float16_t *base, svfloat16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_f16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_f32u10__SVBool_tPf13svfloat32x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_f32(svbool_t pg, float32_t *base, svfloat32x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_f32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst3_f64u10__SVBool_tPd13svfloat64x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_f64(svbool_t pg, float64_t *base, svfloat64x3_t data)
 {
   return SVE_ACLE_FUNC(svst3,_f64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_s8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z18test_svst3_vnum_s8u10__SVBool_tPal10svint8x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_s8,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_s16u10__SVBool_tPsl11svint16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_s16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_s32u10__SVBool_tPil11svint32x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_s32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_s64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_s64u10__SVBool_tPll11svint64x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_s64,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_u8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z18test_svst3_vnum_u8u10__SVBool_tPhl11svuint8x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv48i8(<vscale x 48 x i8> [[DATA]], i64 32)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_u8,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_u16u10__SVBool_tPtl12svuint16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv24i16(<vscale x 24 x i16> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_u16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_u32u10__SVBool_tPjl12svuint32x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv12i32(<vscale x 12 x i32> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_u32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_u64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_u64u10__SVBool_tPml12svuint64x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv6i64(<vscale x 6 x i64> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_u64,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_f16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast half* [[BASE:%.*]] to <vscale x 8 x half>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_f16u10__SVBool_tPDhl13svfloat16x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast half* [[BASE:%.*]] to <vscale x 8 x half>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv24f16(<vscale x 24 x half> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_f16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[BASE:%.*]] to <vscale x 4 x float>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_f32u10__SVBool_tPfl13svfloat32x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[BASE:%.*]] to <vscale x 4 x float>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv12f32(<vscale x 12 x float> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_f32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst3_vnum_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[BASE:%.*]] to <vscale x 2 x double>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst3_vnum_f64u10__SVBool_tPdl13svfloat64x3_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[BASE:%.*]] to <vscale x 2 x double>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i32 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv6f64(<vscale x 6 x double> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst3_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x3_t data)
 {
   return SVE_ACLE_FUNC(svst3_vnum,_f64,,)(pg, base, vnum, data);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st4-bfloat.c

 #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
 
 // CHECK-LABEL: @test_svst4_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z15test_svst4_bf16u10__SVBool_tPu6__bf1614svbfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x bfloat> [[TMP2]], <vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_bf16(svbool_t pg, bfloat16_t *base, svbfloat16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_bf16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_bf16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast bfloat* [[BASE:%.*]] to <vscale x 8 x bfloat>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z20test_svst4_vnum_bf16u10__SVBool_tPu6__bf16l14svbfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast bfloat* [[BASE:%.*]] to <vscale x 8 x bfloat>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv32bf16(<vscale x 32 x bfloat> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8bf16(<vscale x 8 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], <vscale x 8 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], <vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_bf16(svbool_t pg, bfloat16_t *base, int64_t vnum, svbfloat16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_bf16,,)(pg, base, vnum, data);
 }

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_st4.c

 // 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
 
 // CHECK-LABEL: @test_svst4_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z13test_svst4_s8u10__SVBool_tPa10svint8x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_s8(svbool_t pg, int8_t *base, svint8x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_s8,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_s16u10__SVBool_tPs11svint16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_s16(svbool_t pg, int16_t *base, svint16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_s16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_s32u10__SVBool_tPi11svint32x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_s32(svbool_t pg, int32_t *base, svint32x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_s32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_s64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_s64u10__SVBool_tPl11svint64x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_s64(svbool_t pg, int64_t *base, svint64x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_s64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z13test_svst4_u8u10__SVBool_tPh11svuint8x4_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_u8(svbool_t pg, uint8_t *base, svuint8x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_u8,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_u16u10__SVBool_tPt12svuint16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[TMP2]], <vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_u16(svbool_t pg, uint16_t *base, svuint16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_u16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_u32u10__SVBool_tPj12svuint32x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_u32(svbool_t pg, uint32_t *base, svuint32x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_u32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_u64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_u64u10__SVBool_tPm12svuint64x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[TMP2]], <vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_u64(svbool_t pg, uint64_t *base, svuint64x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_u64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_f16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_f16u10__SVBool_tPDh13svfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_f16(svbool_t pg, float16_t *base, svfloat16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_f16,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_f32u10__SVBool_tPf13svfloat32x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP1]], <vscale x 4 x float> [[TMP2]], <vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_f32(svbool_t pg, float32_t *base, svfloat32x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_f32,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z14test_svst4_f64u10__SVBool_tPd13svfloat64x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP1]], <vscale x 2 x double> [[TMP2]], <vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_f64(svbool_t pg, float64_t *base, svfloat64x4_t data)
 {
   return SVE_ACLE_FUNC(svst4,_f64,,)(pg, base, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_s8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z18test_svst4_vnum_s8u10__SVBool_tPal10svint8x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_s8,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_s16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_s16u10__SVBool_tPsl11svint16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_s16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_s32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_s32u10__SVBool_tPil11svint32x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_s32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_s64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_s64u10__SVBool_tPll11svint64x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_s64,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_u8(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z18test_svst4_vnum_u8u10__SVBool_tPhl11svuint8x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[BASE:%.*]] to <vscale x 16 x i8>*
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* [[TMP0]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv64i8(<vscale x 64 x i8> [[DATA]], i64 48)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], <vscale x 16 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], <vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_u8,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_u16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_u16u10__SVBool_tPtl12svuint16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[BASE:%.*]] to <vscale x 8 x i16>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv32i16(<vscale x 32 x i16> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], <vscale x 8 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], <vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_u16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_u32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_u32u10__SVBool_tPjl12svuint32x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[BASE:%.*]] to <vscale x 4 x i32>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv16i32(<vscale x 16 x i32> [[DATA]], i64 12)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], <vscale x 4 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], <vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_u32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_u64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_u64u10__SVBool_tPml12svuint64x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[BASE:%.*]] to <vscale x 2 x i64>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv8i64(<vscale x 8 x i64> [[DATA]], i64 6)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], <vscale x 2 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], <vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_u64,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_f16(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast half* [[BASE:%.*]] to <vscale x 8 x half>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_f16u10__SVBool_tPDhl13svfloat16x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast half* [[BASE:%.*]] to <vscale x 8 x half>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[DATA]], i64 24)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_f16,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_f32(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[BASE:%.*]] to <vscale x 4 x float>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_f32u10__SVBool_tPfl13svfloat32x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[BASE:%.*]] to <vscale x 4 x float>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv16f32(<vscale x 16 x float> [[DATA]], i64 12)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], <vscale x 4 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], <vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_f32,,)(pg, base, vnum, data);
 }
 
 // CHECK-LABEL: @test_svst4_vnum_f64(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[BASE:%.*]] to <vscale x 2 x double>*
 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 1)
-// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 2)
-// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 3)
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
 // CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
 // CHECK-NEXT:    ret void
 //
 // CPP-CHECK-LABEL: @_Z19test_svst4_vnum_f64u10__SVBool_tPdl13svfloat64x4_t(
 // CPP-CHECK-NEXT:  entry:
 // CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
 // CPP-CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[BASE:%.*]] to <vscale x 2 x double>*
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* [[TMP1]], i64 [[VNUM:%.*]], i64 0
-// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 1)
-// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 2)
-// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i32 3)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv8f64(<vscale x 8 x double> [[DATA]], i64 6)
 // CPP-CHECK-NEXT:    tail call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], <vscale x 2 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], <vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
 // CPP-CHECK-NEXT:    ret void
 //
 void test_svst4_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x4_t data)
 {
   return SVE_ACLE_FUNC(svst4_vnum,_f64,,)(pg, base, vnum, data);
 }

clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_tbl2-bfloat.c

 // 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
 
 // CHECK-LABEL: @test_svtbl2_bf16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbl2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z16test_svtbl2_bf1614svbfloat16x2_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tuple.get.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x bfloat> @llvm.vector.extract.nxv8bf16.nxv16bf16(<vscale x 16 x bfloat> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.tbl2.nxv8bf16(<vscale x 8 x bfloat> [[TMP0]], <vscale x 8 x bfloat> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP2]]
 //
 svbfloat16_t test_svtbl2_bf16(svbfloat16x2_t data, svuint16_t indices) {
   return SVE_ACLE_FUNC(svtbl2, _bf16, , )(data, indices);
 }

clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_tbl2.c

 // 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
 
 // CHECK-LABEL: @test_svtbl2_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svtbl2_s810svint8x2_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP2]]
 //
 svint8_t test_svtbl2_s8(svint8x2_t data, svuint8_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_s8,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_s16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_s1611svint16x2_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP2]]
 //
 svint16_t test_svtbl2_s16(svint16x2_t data, svuint16_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_s16,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_s32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_s3211svint32x2_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP2]]
 //
 svint32_t test_svtbl2_s32(svint32x2_t data, svuint32_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_s32,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_s64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_s6411svint64x2_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP2]]
 //
 svint64_t test_svtbl2_s64(svint64x2_t data, svuint64_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_s64,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svtbl2_u811svuint8x2_tu11__SVUint8_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 16 x i8> @llvm.vector.extract.nxv16i8.nxv32i8(<vscale x 32 x i8> [[DATA]], i64 16)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sve.tbl2.nxv16i8(<vscale x 16 x i8> [[TMP0]], <vscale x 16 x i8> [[TMP1]], <vscale x 16 x i8> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP2]]
 //
 svuint8_t test_svtbl2_u8(svuint8x2_t data, svuint8_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_u8,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_u16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_u1612svuint16x2_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x i16> @llvm.vector.extract.nxv8i16.nxv16i16(<vscale x 16 x i16> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sve.tbl2.nxv8i16(<vscale x 8 x i16> [[TMP0]], <vscale x 8 x i16> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP2]]
 //
 svuint16_t test_svtbl2_u16(svuint16x2_t data, svuint16_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_u16,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_u32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_u3212svuint32x2_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x i32> @llvm.vector.extract.nxv4i32.nxv8i32(<vscale x 8 x i32> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.tbl2.nxv4i32(<vscale x 4 x i32> [[TMP0]], <vscale x 4 x i32> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP2]]
 //
 svuint32_t test_svtbl2_u32(svuint32x2_t data, svuint32_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_u32,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_u64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_u6412svuint64x2_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x i64> @llvm.vector.extract.nxv2i64.nxv4i64(<vscale x 4 x i64> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x i64> @llvm.aarch64.sve.tbl2.nxv2i64(<vscale x 2 x i64> [[TMP0]], <vscale x 2 x i64> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP2]]
 //
 svuint64_t test_svtbl2_u64(svuint64x2_t data, svuint64_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_u64,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_f16(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 8 x half> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_f1613svfloat16x2_tu12__SVUint16_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[DATA]], i64 8)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.tbl2.nxv8f16(<vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x i16> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 8 x half> [[TMP2]]
 //
 svfloat16_t test_svtbl2_f16(svfloat16x2_t data, svuint16_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_f16,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_f32(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 4 x float> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_f3213svfloat32x2_tu12__SVUint32_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.vector.extract.nxv4f32.nxv8f32(<vscale x 8 x float> [[DATA]], i64 4)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.tbl2.nxv4f32(<vscale x 4 x float> [[TMP0]], <vscale x 4 x float> [[TMP1]], <vscale x 4 x i32> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 4 x float> [[TMP2]]
 //
 svfloat32_t test_svtbl2_f32(svfloat32x2_t data, svuint32_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_f32,,)(data, indices);
 }
 
 // CHECK-LABEL: @test_svtbl2_f64(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i32 0)
-// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i32 1)
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
 // CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
 // CHECK-NEXT:    ret <vscale x 2 x double> [[TMP2]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svtbl2_f6413svfloat64x2_tu12__SVUint64_t(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i32 0)
-// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i32 1)
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA:%.*]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> [[DATA]], i64 2)
 // CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.tbl2.nxv2f64(<vscale x 2 x double> [[TMP0]], <vscale x 2 x double> [[TMP1]], <vscale x 2 x i64> [[INDICES:%.*]])
 // CPP-CHECK-NEXT:    ret <vscale x 2 x double> [[TMP2]]
 //
 svfloat64_t test_svtbl2_f64(svfloat64x2_t data, svuint64_t indices)
 {
   return SVE_ACLE_FUNC(svtbl2,_f64,,)(data, indices);
 }