diff --git a/clang/include/clang/Basic/arm_sve.td b/clang/include/clang/Basic/arm_sve.td
--- a/clang/include/clang/Basic/arm_sve.td
+++ b/clang/include/clang/Basic/arm_sve.td
@@ -512,14 +512,14 @@
 }
 
 // Load N-element structure into N vectors (scalar base)
-defm SVLD2 : StructLoad<"svld2[_{2}]", "2Pc", "aarch64_sve_ld2">;
-defm SVLD3 : StructLoad<"svld3[_{2}]", "3Pc", "aarch64_sve_ld3">;
-defm SVLD4 : StructLoad<"svld4[_{2}]", "4Pc", "aarch64_sve_ld4">;
+defm SVLD2 : StructLoad<"svld2[_{2}]", "2Pc", "aarch64_sve_ld2_sret">;
+defm SVLD3 : StructLoad<"svld3[_{2}]", "3Pc", "aarch64_sve_ld3_sret">;
+defm SVLD4 : StructLoad<"svld4[_{2}]", "4Pc", "aarch64_sve_ld4_sret">;
 
 // Load N-element structure into N vectors (scalar base, VL displacement)
-defm SVLD2_VNUM : StructLoad<"svld2_vnum[_{2}]", "2Pcl", "aarch64_sve_ld2">;
-defm SVLD3_VNUM : StructLoad<"svld3_vnum[_{2}]", "3Pcl", "aarch64_sve_ld3">;
-defm SVLD4_VNUM : StructLoad<"svld4_vnum[_{2}]", "4Pcl", "aarch64_sve_ld4">;
+defm SVLD2_VNUM : StructLoad<"svld2_vnum[_{2}]", "2Pcl", "aarch64_sve_ld2_sret">;
+defm SVLD3_VNUM : StructLoad<"svld3_vnum[_{2}]", "3Pcl", "aarch64_sve_ld3_sret">;
+defm SVLD4_VNUM : StructLoad<"svld4_vnum[_{2}]", "4Pcl", "aarch64_sve_ld4_sret">;
 
 // Load one octoword and replicate (scalar base)
 let ArchGuard = "defined(__ARM_FEATURE_SVE_MATMUL_FP64)" in {
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -8841,13 +8841,13 @@
 
   unsigned N;
   switch (IntID) {
-  case Intrinsic::aarch64_sve_ld2:
+  case Intrinsic::aarch64_sve_ld2_sret:
     N = 2;
     break;
-  case Intrinsic::aarch64_sve_ld3:
+  case Intrinsic::aarch64_sve_ld3_sret:
     N = 3;
     break;
-  case Intrinsic::aarch64_sve_ld4:
+  case Intrinsic::aarch64_sve_ld4_sret:
     N = 4;
     break;
   default:
@@ -8861,9 +8861,16 @@
   Value *Offset = Ops.size() > 2 ? Ops[2] : Builder.getInt32(0);
   BasePtr = Builder.CreateGEP(VTy, BasePtr, Offset);
   BasePtr = Builder.CreateBitCast(BasePtr, EltPtrTy);
-
-  Function *F = CGM.getIntrinsic(IntID, {RetTy, Predicate->getType()});
-  return Builder.CreateCall(F, { Predicate, BasePtr });
+  Function *F = CGM.getIntrinsic(IntID, {VTy});
+  Value *Call = Builder.CreateCall(F, {Predicate, BasePtr});
+  unsigned MinElts = VTy->getElementCount().getKnownMinValue();
+  Value *Ret = llvm::PoisonValue::get(RetTy);
+  for (unsigned I = 0; I < N; I++) {
+    Value *Idx = ConstantInt::get(CGM.Int64Ty, I * MinElts);
+    Value *SRet = Builder.CreateExtractValue(Call, I);
+    Ret = Builder.CreateInsertVector(RetTy, Ret, SRet, Idx);
+  }
+  return Ret;
 }
 
 Value *CodeGenFunction::EmitSVEStructStore(const SVETypeFlags &TypeFlags,
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2-bfloat.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2-bfloat.c
--- a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2-bfloat.c
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2-bfloat.c
@@ -18,14 +18,22 @@
 // CHECK-LABEL: @test_svld2_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 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svld2_bf16u10__SVBool_tPKu6__bf16(
 // 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 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP5]]
 //
 svbfloat16x2_t test_svld2_bf16(svbool_t pg, const bfloat16_t *base)
 {
@@ -38,16 +46,24 @@
 // 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 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z20test_svld2_vnum_bf16u10__SVBool_tPKu6__bf16l(
 // 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 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> poison, <vscale x 8 x bfloat> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x bfloat> @llvm.vector.insert.nxv16bf16.nxv8bf16(<vscale x 16 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x bfloat> [[TMP7]]
 //
 svbfloat16x2_t test_svld2_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum)
 {
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2.c
--- a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2.c
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld2.c
@@ -17,13 +17,21 @@
 
 // CHECK-LABEL: @test_svld2_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP4]]
 //
 // CPP-CHECK-LABEL: @_Z13test_svld2_s8u10__SVBool_tPKa(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP4]]
 //
 svint8x2_t test_svld2_s8(svbool_t pg, const int8_t *base)
 {
@@ -33,14 +41,22 @@
 // CHECK-LABEL: @test_svld2_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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_s16u10__SVBool_tPKs(
 // 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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP5]]
 //
 svint16x2_t test_svld2_s16(svbool_t pg, const int16_t *base)
 {
@@ -50,14 +66,22 @@
 // CHECK-LABEL: @test_svld2_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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_s32u10__SVBool_tPKi(
 // 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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP5]]
 //
 svint32x2_t test_svld2_s32(svbool_t pg, const int32_t *base)
 {
@@ -67,14 +91,22 @@
 // CHECK-LABEL: @test_svld2_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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_s64u10__SVBool_tPKl(
 // 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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP5]]
 //
 svint64x2_t test_svld2_s64(svbool_t pg, const int64_t *base)
 {
@@ -83,13 +115,21 @@
 
 // CHECK-LABEL: @test_svld2_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP4]]
 //
 // CPP-CHECK-LABEL: @_Z13test_svld2_u8u10__SVBool_tPKh(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP4]]
 //
 svuint8x2_t test_svld2_u8(svbool_t pg, const uint8_t *base)
 {
@@ -99,14 +139,22 @@
 // CHECK-LABEL: @test_svld2_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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_u16u10__SVBool_tPKt(
 // 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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP5]]
 //
 svuint16x2_t test_svld2_u16(svbool_t pg, const uint16_t *base)
 {
@@ -116,14 +164,22 @@
 // CHECK-LABEL: @test_svld2_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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_u32u10__SVBool_tPKj(
 // 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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP5]]
 //
 svuint32x2_t test_svld2_u32(svbool_t pg, const uint32_t *base)
 {
@@ -133,14 +189,22 @@
 // CHECK-LABEL: @test_svld2_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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_u64u10__SVBool_tPKm(
 // 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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP5]]
 //
 svuint64x2_t test_svld2_u64(svbool_t pg, const uint64_t *base)
 {
@@ -150,14 +214,22 @@
 // CHECK-LABEL: @test_svld2_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 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x half> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x half> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_f16u10__SVBool_tPKDh(
 // 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 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP5]]
 //
 svfloat16x2_t test_svld2_f16(svbool_t pg, const float16_t *base)
 {
@@ -167,14 +239,22 @@
 // CHECK-LABEL: @test_svld2_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 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x float> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x float> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_f32u10__SVBool_tPKf(
 // 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 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP5]]
 //
 svfloat32x2_t test_svld2_f32(svbool_t pg, const float32_t *base)
 {
@@ -184,14 +264,22 @@
 // CHECK-LABEL: @test_svld2_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 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 4 x double> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x double> [[TMP5]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld2_f64u10__SVBool_tPKd(
 // 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 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP5]]
 //
 svfloat64x2_t test_svld2_f64(svbool_t pg, const float64_t *base)
 {
@@ -202,15 +290,23 @@
 // 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 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP2]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP6]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svld2_vnum_s8u10__SVBool_tPKal(
 // 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 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP2]]
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP6]]
 //
 svint8x2_t test_svld2_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum)
 {
@@ -222,16 +318,24 @@
 // 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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_s16u10__SVBool_tPKsl(
 // 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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP7]]
 //
 svint16x2_t test_svld2_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
 {
@@ -243,16 +347,24 @@
 // 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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_s32u10__SVBool_tPKil(
 // 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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP7]]
 //
 svint32x2_t test_svld2_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
 {
@@ -264,16 +376,24 @@
 // 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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_s64u10__SVBool_tPKll(
 // 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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP7]]
 //
 svint64x2_t test_svld2_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
 {
@@ -284,15 +404,23 @@
 // 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 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP2]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP6]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svld2_vnum_u8u10__SVBool_tPKhl(
 // 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 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP2]]
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 32 x i8> @llvm.vector.insert.nxv32i8.nxv16i8(<vscale x 32 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i8> [[TMP6]]
 //
 svuint8x2_t test_svld2_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum)
 {
@@ -304,16 +432,24 @@
 // 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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_u16u10__SVBool_tPKtl(
 // 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 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i16> @llvm.vector.insert.nxv16i16.nxv8i16(<vscale x 16 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i16> [[TMP7]]
 //
 svuint16x2_t test_svld2_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum)
 {
@@ -325,16 +461,24 @@
 // 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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_u32u10__SVBool_tPKjl(
 // 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 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i32> @llvm.vector.insert.nxv8i32.nxv4i32(<vscale x 8 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i32> [[TMP7]]
 //
 svuint32x2_t test_svld2_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum)
 {
@@ -346,16 +490,24 @@
 // 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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_u64u10__SVBool_tPKml(
 // 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 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 4 x i64> @llvm.vector.insert.nxv4i64.nxv2i64(<vscale x 4 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i64> [[TMP7]]
 //
 svuint64x2_t test_svld2_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum)
 {
@@ -367,16 +519,24 @@
 // 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 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x half> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 16 x half> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_f16u10__SVBool_tPKDhl(
 // 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 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> poison, <vscale x 8 x half> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x half> @llvm.vector.insert.nxv16f16.nxv8f16(<vscale x 16 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x half> [[TMP7]]
 //
 svfloat16x2_t test_svld2_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum)
 {
@@ -388,16 +548,24 @@
 // 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 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 8 x float> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 4)
+// CHECK-NEXT:    ret <vscale x 8 x float> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_f32u10__SVBool_tPKfl(
 // 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 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> poison, <vscale x 4 x float> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x float> @llvm.vector.insert.nxv8f32.nxv4f32(<vscale x 8 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x float> [[TMP7]]
 //
 svfloat32x2_t test_svld2_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum)
 {
@@ -409,16 +577,24 @@
 // 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 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 4 x double> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 2)
+// CHECK-NEXT:    ret <vscale x 4 x double> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld2_vnum_f64u10__SVBool_tPKdl(
 // 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 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 4 x double> @llvm.vector.insert.nxv4f64.nxv2f64(<vscale x 4 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x double> [[TMP7]]
 //
 svfloat64x2_t test_svld2_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum)
 {
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3-bfloat.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3-bfloat.c
--- a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3-bfloat.c
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3-bfloat.c
@@ -18,14 +18,26 @@
 // CHECK-LABEL: @test_svld3_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 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svld3_bf16u10__SVBool_tPKu6__bf16(
 // 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 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP7]]
 //
 svbfloat16x3_t test_svld3_bf16(svbool_t pg, const bfloat16_t *base)
 {
@@ -37,16 +49,28 @@
 // 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 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z20test_svld3_vnum_bf16u10__SVBool_tPKu6__bf16l(
 // 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 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> poison, <vscale x 8 x bfloat> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x bfloat> @llvm.vector.insert.nxv24bf16.nxv8bf16(<vscale x 24 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x bfloat> [[TMP9]]
 //
 svbfloat16x3_t test_svld3_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum)
 {
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3.c
--- a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3.c
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld3.c
@@ -17,13 +17,25 @@
 
 // CHECK-LABEL: @test_svld3_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP6]]
 //
 // CPP-CHECK-LABEL: @_Z13test_svld3_s8u10__SVBool_tPKa(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP6]]
 //
 svint8x3_t test_svld3_s8(svbool_t pg, const int8_t *base)
 {
@@ -33,14 +45,26 @@
 // CHECK-LABEL: @test_svld3_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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_s16u10__SVBool_tPKs(
 // 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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP7]]
 //
 svint16x3_t test_svld3_s16(svbool_t pg, const int16_t *base)
 {
@@ -50,14 +74,26 @@
 // CHECK-LABEL: @test_svld3_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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_s32u10__SVBool_tPKi(
 // 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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP7]]
 //
 svint32x3_t test_svld3_s32(svbool_t pg, const int32_t *base)
 {
@@ -67,14 +103,26 @@
 // CHECK-LABEL: @test_svld3_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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_s64u10__SVBool_tPKl(
 // 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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP7]]
 //
 svint64x3_t test_svld3_s64(svbool_t pg, const int64_t *base)
 {
@@ -83,13 +131,25 @@
 
 // CHECK-LABEL: @test_svld3_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP6]]
 //
 // CPP-CHECK-LABEL: @_Z13test_svld3_u8u10__SVBool_tPKh(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP6]]
 //
 svuint8x3_t test_svld3_u8(svbool_t pg, const uint8_t *base)
 {
@@ -99,14 +159,26 @@
 // CHECK-LABEL: @test_svld3_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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_u16u10__SVBool_tPKt(
 // 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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP7]]
 //
 svuint16x3_t test_svld3_u16(svbool_t pg, const uint16_t *base)
 {
@@ -116,14 +188,26 @@
 // CHECK-LABEL: @test_svld3_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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_u32u10__SVBool_tPKj(
 // 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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP7]]
 //
 svuint32x3_t test_svld3_u32(svbool_t pg, const uint32_t *base)
 {
@@ -133,14 +217,26 @@
 // CHECK-LABEL: @test_svld3_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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_u64u10__SVBool_tPKm(
 // 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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP7]]
 //
 svuint64x3_t test_svld3_u64(svbool_t pg, const uint64_t *base)
 {
@@ -150,14 +246,26 @@
 // CHECK-LABEL: @test_svld3_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 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 24 x half> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x half> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_f16u10__SVBool_tPKDh(
 // 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 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP7]]
 //
 svfloat16x3_t test_svld3_f16(svbool_t pg, const float16_t *base)
 {
@@ -167,14 +275,26 @@
 // CHECK-LABEL: @test_svld3_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 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 12 x float> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x float> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_f32u10__SVBool_tPKf(
 // 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 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP7]]
 //
 svfloat32x3_t test_svld3_f32(svbool_t pg, const float32_t *base)
 {
@@ -184,14 +304,26 @@
 // CHECK-LABEL: @test_svld3_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 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 6 x double> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x double> [[TMP7]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld3_f64u10__SVBool_tPKd(
 // 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 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP7]]
 //
 svfloat64x3_t test_svld3_f64(svbool_t pg, const float64_t *base)
 {
@@ -202,15 +334,27 @@
 // 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 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP8]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svld3_vnum_s8u10__SVBool_tPKal(
 // 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 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP8]]
 //
 svint8x3_t test_svld3_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum)
 {
@@ -222,16 +366,28 @@
 // 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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_s16u10__SVBool_tPKsl(
 // 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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP9]]
 //
 svint16x3_t test_svld3_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
 {
@@ -243,16 +399,28 @@
 // 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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_s32u10__SVBool_tPKil(
 // 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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP9]]
 //
 svint32x3_t test_svld3_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
 {
@@ -264,16 +432,28 @@
 // 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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_s64u10__SVBool_tPKll(
 // 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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP9]]
 //
 svint64x3_t test_svld3_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
 {
@@ -284,15 +464,27 @@
 // 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 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP8]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svld3_vnum_u8u10__SVBool_tPKhl(
 // 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 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP2]]
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 48 x i8> @llvm.vector.insert.nxv48i8.nxv16i8(<vscale x 48 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CPP-CHECK-NEXT:    ret <vscale x 48 x i8> [[TMP8]]
 //
 svuint8x3_t test_svld3_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum)
 {
@@ -304,16 +496,28 @@
 // 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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_u16u10__SVBool_tPKtl(
 // 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 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x i16> @llvm.vector.insert.nxv24i16.nxv8i16(<vscale x 24 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x i16> [[TMP9]]
 //
 svuint16x3_t test_svld3_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum)
 {
@@ -325,16 +529,28 @@
 // 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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_u32u10__SVBool_tPKjl(
 // 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 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 12 x i32> @llvm.vector.insert.nxv12i32.nxv4i32(<vscale x 12 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x i32> [[TMP9]]
 //
 svuint32x3_t test_svld3_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum)
 {
@@ -346,16 +562,28 @@
 // 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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_u64u10__SVBool_tPKml(
 // 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 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 6 x i64> @llvm.vector.insert.nxv6i64.nxv2i64(<vscale x 6 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x i64> [[TMP9]]
 //
 svuint64x3_t test_svld3_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum)
 {
@@ -367,16 +595,28 @@
 // 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 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 24 x half> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 16)
+// CHECK-NEXT:    ret <vscale x 24 x half> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_f16u10__SVBool_tPKDhl(
 // 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 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> poison, <vscale x 8 x half> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 24 x half> @llvm.vector.insert.nxv24f16.nxv8f16(<vscale x 24 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    ret <vscale x 24 x half> [[TMP9]]
 //
 svfloat16x3_t test_svld3_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum)
 {
@@ -388,16 +628,28 @@
 // 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 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 12 x float> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 8)
+// CHECK-NEXT:    ret <vscale x 12 x float> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_f32u10__SVBool_tPKfl(
 // 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 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> poison, <vscale x 4 x float> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 12 x float> @llvm.vector.insert.nxv12f32.nxv4f32(<vscale x 12 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 8)
+// CPP-CHECK-NEXT:    ret <vscale x 12 x float> [[TMP9]]
 //
 svfloat32x3_t test_svld3_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum)
 {
@@ -409,16 +661,28 @@
 // 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 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 6 x double> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 2)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 4)
+// CHECK-NEXT:    ret <vscale x 6 x double> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld3_vnum_f64u10__SVBool_tPKdl(
 // 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 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nxv2f64(<vscale x 6 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 4)
+// CPP-CHECK-NEXT:    ret <vscale x 6 x double> [[TMP9]]
 //
 svfloat64x3_t test_svld3_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum)
 {
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4-bfloat.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4-bfloat.c
--- a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4-bfloat.c
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4-bfloat.c
@@ -18,14 +18,30 @@
 // CHECK-LABEL: @test_svld4_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 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z15test_svld4_bf16u10__SVBool_tPKu6__bf16(
 // 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 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP3]], <vscale x 8 x bfloat> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP9]]
 //
 svbfloat16x4_t test_svld4_bf16(svbool_t pg, const bfloat16_t *base)
 {
@@ -37,16 +53,32 @@
 // 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 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 16)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP9]], <vscale x 8 x bfloat> [[TMP10]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z20test_svld4_vnum_bf16u10__SVBool_tPKu6__bf16l(
 // 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 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], bfloat* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> poison, <vscale x 8 x bfloat> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP5]], <vscale x 8 x bfloat> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP7]], <vscale x 8 x bfloat> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x bfloat> @llvm.vector.insert.nxv32bf16.nxv8bf16(<vscale x 32 x bfloat> [[TMP9]], <vscale x 8 x bfloat> [[TMP10]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x bfloat> [[TMP11]]
 //
 svbfloat16x4_t test_svld4_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum)
 {
diff --git a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4.c b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4.c
--- a/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4.c
+++ b/clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_ld4.c
@@ -17,13 +17,29 @@
 
 // CHECK-LABEL: @test_svld4_s8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
+// CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
+// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP8]]
 //
 // CPP-CHECK-LABEL: @_Z13test_svld4_s8u10__SVBool_tPKa(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
+// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP8]]
 //
 svint8x4_t test_svld4_s8(svbool_t pg, const int8_t *base)
 {
@@ -33,14 +49,30 @@
 // CHECK-LABEL: @test_svld4_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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_s16u10__SVBool_tPKs(
 // 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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP9]]
 //
 svint16x4_t test_svld4_s16(svbool_t pg, const int16_t *base)
 {
@@ -50,14 +82,30 @@
 // CHECK-LABEL: @test_svld4_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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_s32u10__SVBool_tPKi(
 // 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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP9]]
 //
 svint32x4_t test_svld4_s32(svbool_t pg, const int32_t *base)
 {
@@ -67,14 +115,30 @@
 // CHECK-LABEL: @test_svld4_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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_s64u10__SVBool_tPKl(
 // 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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP9]]
 //
 svint64x4_t test_svld4_s64(svbool_t pg, const int64_t *base)
 {
@@ -83,13 +147,29 @@
 
 // CHECK-LABEL: @test_svld4_u8(
 // CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
+// CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
+// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP8]]
 //
 // CPP-CHECK-LABEL: @_Z13test_svld4_u8u10__SVBool_tPKh(
 // CPP-CHECK-NEXT:  entry:
-// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP0]]
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 0
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP1]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 1
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP2]], <vscale x 16 x i8> [[TMP3]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 2
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]], 3
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 48)
+// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP8]]
 //
 svuint8x4_t test_svld4_u8(svbool_t pg, const uint8_t *base)
 {
@@ -99,14 +179,30 @@
 // CHECK-LABEL: @test_svld4_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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_u16u10__SVBool_tPKt(
 // 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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP3]], <vscale x 8 x i16> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP9]]
 //
 svuint16x4_t test_svld4_u16(svbool_t pg, const uint16_t *base)
 {
@@ -116,14 +212,30 @@
 // CHECK-LABEL: @test_svld4_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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_u32u10__SVBool_tPKj(
 // 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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP3]], <vscale x 4 x i32> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP9]]
 //
 svuint32x4_t test_svld4_u32(svbool_t pg, const uint32_t *base)
 {
@@ -133,14 +245,30 @@
 // CHECK-LABEL: @test_svld4_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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_u64u10__SVBool_tPKm(
 // 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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP3]], <vscale x 2 x i64> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP9]]
 //
 svuint64x4_t test_svld4_u64(svbool_t pg, const uint64_t *base)
 {
@@ -150,14 +278,30 @@
 // CHECK-LABEL: @test_svld4_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 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 32 x half> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x half> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_f16u10__SVBool_tPKDh(
 // 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 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP9]]
 //
 svfloat16x4_t test_svld4_f16(svbool_t pg, const float16_t *base)
 {
@@ -167,14 +311,30 @@
 // CHECK-LABEL: @test_svld4_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 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 16 x float> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x float> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_f32u10__SVBool_tPKf(
 // 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 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP3]], <vscale x 4 x float> [[TMP4]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP9]]
 //
 svfloat32x4_t test_svld4_f32(svbool_t pg, const float32_t *base)
 {
@@ -184,14 +344,30 @@
 // CHECK-LABEL: @test_svld4_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 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
-// CHECK-NEXT:    ret <vscale x 8 x double> [[TMP1]]
+// CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
+// CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 3
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x double> [[TMP9]]
 //
 // CPP-CHECK-LABEL: @_Z14test_svld4_f64u10__SVBool_tPKd(
 // 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 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP1]]
+// CPP-CHECK-NEXT:    [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[BASE:%.*]])
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 0
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP2]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 1
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP3]], <vscale x 2 x double> [[TMP4]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 2
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]], 3
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP9]]
 //
 svfloat64x4_t test_svld4_f64(svbool_t pg, const float64_t *base)
 {
@@ -202,15 +378,31 @@
 // 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 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP2]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 3
+// CHECK-NEXT:    [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 48)
+// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP10]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svld4_vnum_s8u10__SVBool_tPKal(
 // 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 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP2]]
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 3
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 48)
+// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP10]]
 //
 svint8x4_t test_svld4_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum)
 {
@@ -222,16 +414,32 @@
 // 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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP9]], <vscale x 8 x i16> [[TMP10]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_s16u10__SVBool_tPKsl(
 // 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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP9]], <vscale x 8 x i16> [[TMP10]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP11]]
 //
 svint16x4_t test_svld4_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
 {
@@ -243,16 +451,32 @@
 // 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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP9]], <vscale x 4 x i32> [[TMP10]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_s32u10__SVBool_tPKil(
 // 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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP9]], <vscale x 4 x i32> [[TMP10]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP11]]
 //
 svint32x4_t test_svld4_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
 {
@@ -264,16 +488,32 @@
 // 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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP9]], <vscale x 2 x i64> [[TMP10]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_s64u10__SVBool_tPKll(
 // 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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP9]], <vscale x 2 x i64> [[TMP10]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP11]]
 //
 svint64x4_t test_svld4_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
 {
@@ -284,15 +524,31 @@
 // 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 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP2]]
+// CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 3
+// CHECK-NEXT:    [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 48)
+// CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP10]]
 //
 // CPP-CHECK-LABEL: @_Z18test_svld4_vnum_u8u10__SVBool_tPKhl(
 // 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 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
-// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP2]]
+// CPP-CHECK-NEXT:    [[TMP2:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], i8* [[TMP1]])
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 0
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> poison, <vscale x 16 x i8> [[TMP3]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 1
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP4]], <vscale x 16 x i8> [[TMP5]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 2
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP6]], <vscale x 16 x i8> [[TMP7]], i64 32)
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP2]], 3
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = tail call <vscale x 64 x i8> @llvm.vector.insert.nxv64i8.nxv16i8(<vscale x 64 x i8> [[TMP8]], <vscale x 16 x i8> [[TMP9]], i64 48)
+// CPP-CHECK-NEXT:    ret <vscale x 64 x i8> [[TMP10]]
 //
 svuint8x4_t test_svld4_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum)
 {
@@ -304,16 +560,32 @@
 // 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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP9]], <vscale x 8 x i16> [[TMP10]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_u16u10__SVBool_tPKtl(
 // 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 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], i16* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> poison, <vscale x 8 x i16> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP5]], <vscale x 8 x i16> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP7]], <vscale x 8 x i16> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x i16> @llvm.vector.insert.nxv32i16.nxv8i16(<vscale x 32 x i16> [[TMP9]], <vscale x 8 x i16> [[TMP10]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x i16> [[TMP11]]
 //
 svuint16x4_t test_svld4_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum)
 {
@@ -325,16 +597,32 @@
 // 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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP9]], <vscale x 4 x i32> [[TMP10]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_u32u10__SVBool_tPKjl(
 // 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 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], i32* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> poison, <vscale x 4 x i32> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP5]], <vscale x 4 x i32> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP7]], <vscale x 4 x i32> [[TMP8]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 16 x i32> @llvm.vector.insert.nxv16i32.nxv4i32(<vscale x 16 x i32> [[TMP9]], <vscale x 4 x i32> [[TMP10]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x i32> [[TMP11]]
 //
 svuint32x4_t test_svld4_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum)
 {
@@ -346,16 +634,32 @@
 // 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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP9]], <vscale x 2 x i64> [[TMP10]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_u64u10__SVBool_tPKml(
 // 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 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], i64* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> poison, <vscale x 2 x i64> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP5]], <vscale x 2 x i64> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP7]], <vscale x 2 x i64> [[TMP8]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 8 x i64> @llvm.vector.insert.nxv8i64.nxv2i64(<vscale x 8 x i64> [[TMP9]], <vscale x 2 x i64> [[TMP10]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i64> [[TMP11]]
 //
 svuint64x4_t test_svld4_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum)
 {
@@ -367,16 +671,32 @@
 // 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 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 32 x half> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 8)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 16)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP9]], <vscale x 8 x half> [[TMP10]], i64 24)
+// CHECK-NEXT:    ret <vscale x 32 x half> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_f16u10__SVBool_tPKDhl(
 // 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 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], half* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> poison, <vscale x 8 x half> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP7]], <vscale x 8 x half> [[TMP8]], i64 16)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 32 x half> @llvm.vector.insert.nxv32f16.nxv8f16(<vscale x 32 x half> [[TMP9]], <vscale x 8 x half> [[TMP10]], i64 24)
+// CPP-CHECK-NEXT:    ret <vscale x 32 x half> [[TMP11]]
 //
 svfloat16x4_t test_svld4_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum)
 {
@@ -388,16 +708,32 @@
 // 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 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 16 x float> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 4)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 8)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP9]], <vscale x 4 x float> [[TMP10]], i64 12)
+// CHECK-NEXT:    ret <vscale x 16 x float> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_f32u10__SVBool_tPKfl(
 // 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 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], float* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> poison, <vscale x 4 x float> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP5]], <vscale x 4 x float> [[TMP6]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP7]], <vscale x 4 x float> [[TMP8]], i64 8)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 16 x float> @llvm.vector.insert.nxv16f32.nxv4f32(<vscale x 16 x float> [[TMP9]], <vscale x 4 x float> [[TMP10]], i64 12)
+// CPP-CHECK-NEXT:    ret <vscale x 16 x float> [[TMP11]]
 //
 svfloat32x4_t test_svld4_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum)
 {
@@ -409,16 +745,32 @@
 // 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 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
-// CHECK-NEXT:    ret <vscale x 8 x double> [[TMP3]]
+// CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
+// CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
+// CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+// CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
+// CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 2)
+// CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
+// CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 4)
+// CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 3
+// CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP9]], <vscale x 2 x double> [[TMP10]], i64 6)
+// CHECK-NEXT:    ret <vscale x 8 x double> [[TMP11]]
 //
 // CPP-CHECK-LABEL: @_Z19test_svld4_vnum_f64u10__SVBool_tPKdl(
 // 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 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
-// CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP3]]
+// CPP-CHECK-NEXT:    [[TMP3:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], double* [[TMP2]])
+// CPP-CHECK-NEXT:    [[TMP4:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 0
+// CPP-CHECK-NEXT:    [[TMP5:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> poison, <vscale x 2 x double> [[TMP4]], i64 0)
+// CPP-CHECK-NEXT:    [[TMP6:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 1
+// CPP-CHECK-NEXT:    [[TMP7:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP5]], <vscale x 2 x double> [[TMP6]], i64 2)
+// CPP-CHECK-NEXT:    [[TMP8:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 2
+// CPP-CHECK-NEXT:    [[TMP9:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP7]], <vscale x 2 x double> [[TMP8]], i64 4)
+// CPP-CHECK-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP3]], 3
+// CPP-CHECK-NEXT:    [[TMP11:%.*]] = tail call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nxv2f64(<vscale x 8 x double> [[TMP9]], <vscale x 2 x double> [[TMP10]], i64 6)
+// CPP-CHECK-NEXT:    ret <vscale x 8 x double> [[TMP11]]
 //
 svfloat64x4_t test_svld4_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum)
 {
diff --git a/llvm/include/llvm/IR/IntrinsicsAArch64.td b/llvm/include/llvm/IR/IntrinsicsAArch64.td
--- a/llvm/include/llvm/IR/IntrinsicsAArch64.td
+++ b/llvm/include/llvm/IR/IntrinsicsAArch64.td
@@ -946,10 +946,6 @@
 
 let TargetPrefix = "aarch64" in {  // All intrinsics start with "llvm.aarch64.".
 
-  class AdvSIMD_ManyVec_PredLoad_Intrinsic
-    : DefaultAttrsIntrinsic<[llvm_anyvector_ty], [llvm_anyvector_ty, LLVMPointerToElt<0>],
-                [IntrReadMem, IntrArgMemOnly]>;
-
   class AdvSIMD_1Vec_PredLoad_Intrinsic
     : DefaultAttrsIntrinsic<[llvm_anyvector_ty],
                 [LLVMScalarOrSameVectorWidth<0, llvm_i1_ty>,
@@ -1529,10 +1525,6 @@
 
 def int_aarch64_sve_ld1   : AdvSIMD_1Vec_PredLoad_Intrinsic;
 
-def int_aarch64_sve_ld2 : AdvSIMD_ManyVec_PredLoad_Intrinsic;
-def int_aarch64_sve_ld3 : AdvSIMD_ManyVec_PredLoad_Intrinsic;
-def int_aarch64_sve_ld4 : AdvSIMD_ManyVec_PredLoad_Intrinsic;
-
 def int_aarch64_sve_ld2_sret : AdvSIMD_2Vec_PredLoad_Intrinsic;
 def int_aarch64_sve_ld3_sret : AdvSIMD_3Vec_PredLoad_Intrinsic;
 def int_aarch64_sve_ld4_sret : AdvSIMD_4Vec_PredLoad_Intrinsic;
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -1029,8 +1029,6 @@
   SDValue LowerWindowsDYNAMIC_STACKALLOC(SDValue Op, SDValue Chain,
                                          SDValue &Size,
                                          SelectionDAG &DAG) const;
-  SDValue LowerSVEStructLoad(unsigned Intrinsic, ArrayRef<SDValue> LoadOps,
-                             EVT VT, SelectionDAG &DAG, const SDLoc &DL) const;
 
   SDValue LowerFixedLengthVectorIntDivideToSVE(SDValue Op,
                                                SelectionDAG &DAG) const;
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -13403,61 +13403,6 @@
   return true;
 }
 
-// Lower an SVE structured load intrinsic returning a tuple type to target
-// specific intrinsic taking the same input but returning a multi-result value
-// of the split tuple type.
-//
-// E.g. Lowering an LD3:
-//
-//  call <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32(
-//                                                    <vscale x 4 x i1> %pred,
-//                                                    <vscale x 4 x i32>* %addr)
-//
-//  Output DAG:
-//
-//    t0: ch = EntryToken
-//        t2: nxv4i1,ch = CopyFromReg t0, Register:nxv4i1 %0
-//        t4: i64,ch = CopyFromReg t0, Register:i64 %1
-//    t5: nxv4i32,nxv4i32,nxv4i32,ch = AArch64ISD::SVE_LD3 t0, t2, t4
-//    t6: nxv12i32 = concat_vectors t5, t5:1, t5:2
-//
-// This is called pre-legalization to avoid widening/splitting issues with
-// non-power-of-2 tuple types used for LD3, such as nxv12i32.
-SDValue AArch64TargetLowering::LowerSVEStructLoad(unsigned Intrinsic,
-                                                  ArrayRef<SDValue> LoadOps,
-                                                  EVT VT, SelectionDAG &DAG,
-                                                  const SDLoc &DL) const {
-  assert(VT.isScalableVector() && "Can only lower scalable vectors");
-
-  unsigned N, Opcode;
-  static const std::pair<unsigned, std::pair<unsigned, unsigned>>
-      IntrinsicMap[] = {
-          {Intrinsic::aarch64_sve_ld2, {2, AArch64ISD::SVE_LD2_MERGE_ZERO}},
-          {Intrinsic::aarch64_sve_ld3, {3, AArch64ISD::SVE_LD3_MERGE_ZERO}},
-          {Intrinsic::aarch64_sve_ld4, {4, AArch64ISD::SVE_LD4_MERGE_ZERO}}};
-
-  std::tie(N, Opcode) = llvm::find_if(IntrinsicMap, [&](auto P) {
-                          return P.first == Intrinsic;
-                        })->second;
-  assert(VT.getVectorElementCount().getKnownMinValue() % N == 0 &&
-         "invalid tuple vector type!");
-
-  EVT SplitVT =
-      EVT::getVectorVT(*DAG.getContext(), VT.getVectorElementType(),
-                       VT.getVectorElementCount().divideCoefficientBy(N));
-  assert(isTypeLegal(SplitVT));
-
-  SmallVector<EVT, 5> VTs(N, SplitVT);
-  VTs.push_back(MVT::Other); // Chain
-  SDVTList NodeTys = DAG.getVTList(VTs);
-
-  SDValue PseudoLoad = DAG.getNode(Opcode, DL, NodeTys, LoadOps);
-  SmallVector<SDValue, 4> PseudoLoadOps;
-  for (unsigned I = 0; I < N; ++I)
-    PseudoLoadOps.push_back(SDValue(PseudoLoad.getNode(), I));
-  return DAG.getNode(ISD::CONCAT_VECTORS, DL, VT, PseudoLoadOps);
-}
-
 EVT AArch64TargetLowering::getOptimalMemOpType(
     const MemOp &Op, const AttributeList &FuncAttributes) const {
   bool CanImplicitFloat = !FuncAttributes.hasFnAttr(Attribute::NoImplicitFloat);
@@ -19852,20 +19797,6 @@
                                         /*OnlyPackedOffsets=*/false);
     case Intrinsic::aarch64_sve_st1_scatter_scalar_offset:
       return performScatterStoreCombine(N, DAG, AArch64ISD::SST1_IMM_PRED);
-    case Intrinsic::aarch64_sve_ld2:
-    case Intrinsic::aarch64_sve_ld3:
-    case Intrinsic::aarch64_sve_ld4: {
-      SDLoc DL(N);
-      SDValue Chain = N->getOperand(0);
-      SDValue Mask = N->getOperand(2);
-      SDValue BasePtr = N->getOperand(3);
-      SDValue LoadOps[] = {Chain, Mask, BasePtr};
-      unsigned IntrinsicID =
-          cast<ConstantSDNode>(N->getOperand(1))->getZExtValue();
-      SDValue Result =
-          LowerSVEStructLoad(IntrinsicID, LoadOps, N->getValueType(0), DAG, DL);
-      return DAG.getMergeValues({Result, Chain}, DL);
-    }
     case Intrinsic::aarch64_rndr:
     case Intrinsic::aarch64_rndrrs: {
       unsigned IntrinsicID =
diff --git a/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll b/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll
--- a/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll
+++ b/llvm/test/CodeGen/AArch64/sve-calling-convention-mixed.ll
@@ -29,10 +29,26 @@
 entry:
   %0 = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
   %1 = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %0)
-  %2 = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> %1, double* %x0)
-  %3 = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> %1, double* %x1)
+  %2 = call {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> %1, double* %x0)
+  %3 = call {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> %1, double* %x1)
   %4 = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %1, double* %x2)
-  %call = call float @callee1(float 1.000000e+00, <vscale x 8 x double> %2, <vscale x 8 x double> %3, <vscale x 2 x double> %4)
+  %5 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  0
+  %6 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  1
+  %7 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  2
+  %8 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  3
+  %9 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> poison, <vscale x 2 x double> %5, i64 0)
+  %10 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %9, <vscale x 2 x double> %6, i64 2)
+  %11 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %10, <vscale x 2 x double> %7, i64 4)
+  %12 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %11, <vscale x 2 x double> %8, i64 6)
+  %13 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %3,  0
+  %14 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %3,  1
+  %15 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %3,  2
+  %16 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %3,  3
+  %17 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> poison, <vscale x 2 x double> %13, i64 0)
+  %18 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %17, <vscale x 2 x double> %14, i64 2)
+  %19 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %18, <vscale x 2 x double> %15, i64 4)
+  %20 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %19, <vscale x 2 x double> %16, i64 6)
+  %call = call float @callee1(float 1.000000e+00, <vscale x 8 x double> %12, <vscale x 8 x double> %20, <vscale x 2 x double> %4)
   ret float %call
 }
 
@@ -52,18 +68,18 @@
 ; CHECK-NEXT:    mov w0, wzr
 ; CHECK-NEXT:    mov w1, #1
 ; CHECK-NEXT:    mov w2, #2
-; CHECK-NEXT:    st1d { z16.d }, p0, [x9]
+; CHECK-NEXT:    st1d { z1.d }, p0, [x9]
 ; CHECK-NEXT:    add x9, sp, #16
 ; CHECK-NEXT:    mov w3, #3
 ; CHECK-NEXT:    mov w4, #4
 ; CHECK-NEXT:    mov w5, #5
 ; CHECK-NEXT:    mov w6, #6
-; CHECK-NEXT:    st1d { z17.d }, p0, [x9, #1, mul vl]
+; CHECK-NEXT:    st1d { z2.d }, p0, [x9, #1, mul vl]
 ; CHECK-NEXT:    add x9, sp, #16
 ; CHECK-NEXT:    mov w7, #7
-; CHECK-NEXT:    st1d { z18.d }, p0, [x9, #2, mul vl]
+; CHECK-NEXT:    st1d { z3.d }, p0, [x9, #2, mul vl]
 ; CHECK-NEXT:    add x9, sp, #16
-; CHECK-NEXT:    st1d { z19.d }, p0, [x9, #3, mul vl]
+; CHECK-NEXT:    st1d { z4.d }, p0, [x9, #3, mul vl]
 ; CHECK-NEXT:    str x8, [sp]
 ; CHECK-NEXT:    bl callee2
 ; CHECK-NEXT:    addvl sp, sp, #4
@@ -73,9 +89,25 @@
 entry:
   %0 = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
   %1 = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %0)
-  %2 = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> %1, double* %x0)
-  %3 = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> %1, double* %x1)
-  %call = call float @callee2(i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, float 1.000000e+00, <vscale x 8 x double> %2, <vscale x 8 x double> %3)
+  %2 = call {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> %1, double* %x0)
+  %3 = call {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> %1, double* %x1)
+  %4 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  0
+  %5 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  1
+  %6 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  2
+  %7 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  3
+  %8 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> poison, <vscale x 2 x double> %4, i64 0)
+  %9 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %8, <vscale x 2 x double> %5, i64 2)
+  %10 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %9, <vscale x 2 x double> %6, i64 4)
+  %11 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %10, <vscale x 2 x double> %7, i64 6)
+  %12 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  0
+  %13 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  1
+  %14 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  2
+  %15 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  3
+  %16 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> poison, <vscale x 2 x double> %12, i64 0)
+  %17 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %16, <vscale x 2 x double> %13, i64 2)
+  %18 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %17, <vscale x 2 x double> %14, i64 4)
+  %19 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %18, <vscale x 2 x double> %15, i64 6)
+  %call = call float @callee2(i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, float 1.000000e+00, <vscale x 8 x double> %11, <vscale x 8 x double> %19)
   ret float %call
 }
 
@@ -92,9 +124,9 @@
 ; CHECK-NEXT:    fmov s1, #2.00000000
 ; CHECK-NEXT:    mov x0, sp
 ; CHECK-NEXT:    ptrue p0.d
-; CHECK-NEXT:    st1d { z16.d }, p0, [sp]
-; CHECK-NEXT:    st1d { z17.d }, p0, [sp, #1, mul vl]
-; CHECK-NEXT:    st1d { z18.d }, p0, [sp, #2, mul vl]
+; CHECK-NEXT:    st1d { z2.d }, p0, [sp]
+; CHECK-NEXT:    st1d { z2.d }, p0, [sp, #1, mul vl]
+; CHECK-NEXT:    st1d { z2.d }, p0, [sp, #2, mul vl]
 ; CHECK-NEXT:    bl callee3
 ; CHECK-NEXT:    addvl sp, sp, #3
 ; CHECK-NEXT:    ldp x29, x30, [sp], #16 // 16-byte Folded Reload
@@ -102,10 +134,24 @@
 entry:
   %0 = call <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 31)
   %1 = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %0)
-  %2 = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1> %1, double* %x0)
-  %3 = call <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1(<vscale x 2 x i1> %1, double* %x1)
-  %4 = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %1, double* %x2)
-  %call = call float @callee3(float 1.000000e+00, float 2.000000e+00, <vscale x 8 x double> %2, <vscale x 6 x double> %3, <vscale x 2 x double> %4)
+  %2 = call {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1> %1, double* %x0)
+  %3 = call {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1> %1, double* %x1)
+  %4 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  0
+  %5 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  1
+  %6 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  2
+  %7 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %2,  3
+  %8 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> poison, <vscale x 2 x double> %4, i64 0)
+  %9 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %8, <vscale x 2 x double> %5, i64 2)
+  %10 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %9, <vscale x 2 x double> %6, i64 4)
+  %11 = call <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double> %10, <vscale x 2 x double> %7, i64 6)
+  %12 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} %3,  0
+  %13 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %3,  1
+  %14 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %3,  2
+  %15 = call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nx2f64(<vscale x 6 x double> poison, <vscale x 2 x double> %4, i64 0)
+  %16 = call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nx2f64(<vscale x 6 x double> %15 , <vscale x 2 x double> %4, i64 2)
+  %17 = call <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nx2f64(<vscale x 6 x double> %16 , <vscale x 2 x double> %4, i64 4)
+  %18 = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %1, double* %x2)
+  %call = call float @callee3(float 1.000000e+00, float 2.000000e+00, <vscale x 8 x double> %11, <vscale x 6 x double> %17, <vscale x 2 x double> %18)
   ret float %call
 }
 
@@ -381,9 +427,9 @@
 
 declare <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32 immarg)
 declare <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1>)
-declare <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1(<vscale x 2 x i1>, double*)
-declare <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1(<vscale x 2 x i1>, double*)
+declare {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld4.sret.nxv2f64(<vscale x 2 x i1>, double*)
+declare {<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>} @llvm.aarch64.sve.ld3.sret.nxv2f64(<vscale x 2 x i1>, double*)
 declare <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1>, double*)
 declare double @llvm.aarch64.sve.faddv.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>)
-declare <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double>, i32 immarg)
-declare <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double>, i32 immarg)
+declare <vscale x 8 x double> @llvm.vector.insert.nxv8f64.nx2f64(<vscale x 8 x double>, <vscale x 2 x double>, i64)
+declare <vscale x 6 x double> @llvm.vector.insert.nxv6f64.nx2f64(<vscale x 6 x double>, <vscale x 2 x double>, i64)
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-ldN-reg+imm-addr-mode.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-ldN-reg+imm-addr-mode.ll
deleted file mode 100644
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-ldN-reg+imm-addr-mode.ll
+++ /dev/null
@@ -1,539 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=aarch64--linux-gnu -mattr=sve < %s | FileCheck %s
-; RUN: llc -mtriple=aarch64--linux-gnu -mattr=sme < %s | FileCheck %s
-
-; NOTE: invalid, upper and lower bound immediate values of the regimm
-; addressing mode are checked only for the byte version of each
-; instruction (`ld<N>b`), as the code for detecting the immediate is
-; common to all instructions, and varies only for the number of
-; elements of the structure store, which is <N> = 2, 3, 4.
-
-; ld2b
-define <vscale x 32 x i8> @ld2.nxv32i8(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld2.nxv32i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, #2, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 2
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8*
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 32 x i8> %res
-}
-
-define <vscale x 32 x i8> @ld2.nxv32i8_lower_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld2.nxv32i8_lower_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, #-16, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 -16
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 32 x i8> %res
-}
-
-define <vscale x 32 x i8> @ld2.nxv32i8_upper_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld2.nxv32i8_upper_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, #14, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 14
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 32 x i8> %res
-}
-
-define <vscale x 32 x i8> @ld2.nxv32i8_not_multiple_of_2(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld2.nxv32i8_not_multiple_of_2:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #3
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 3
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 32 x i8> %res
-}
-
-define <vscale x 32 x i8> @ld2.nxv32i8_outside_lower_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld2.nxv32i8_outside_lower_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #-18
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 -18
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 32 x i8> %res
-}
-
-define <vscale x 32 x i8> @ld2.nxv32i8_outside_upper_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld2.nxv32i8_outside_upper_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #16
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 16
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 32 x i8> %res
-}
-
-; ld2h
-define <vscale x 16 x i16> @ld2.nxv16i16(<vscale x 8 x i1> %Pg, <vscale x 8 x i16>* %addr) {
-; CHECK-LABEL: ld2.nxv16i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, #14, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %addr, i64 14
-%base_ptr = bitcast <vscale x 8 x i16>* %base to i16 *
-%res = call <vscale x 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1.p0i16(<vscale x 8 x i1> %Pg, i16 *%base_ptr)
-ret <vscale x 16 x i16> %res
-}
-
-define <vscale x 16 x half> @ld2.nxv16f16(<vscale x 8 x i1> %Pg, <vscale x 8 x half>* %addr) {
-; CHECK-LABEL: ld2.nxv16f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, #-16, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %addr, i64 -16
-%base_ptr = bitcast <vscale x 8 x half>* %base to half *
-%res = call <vscale x 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1.p0f16(<vscale x 8 x i1> %Pg, half *%base_ptr)
-ret <vscale x 16 x half> %res
-}
-
-define <vscale x 16 x bfloat> @ld2.nxv16bf16(<vscale x 8 x i1> %Pg, <vscale x 8 x bfloat>* %addr) #0 {
-; CHECK-LABEL: ld2.nxv16bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, #12, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* %addr, i64 12
-%base_ptr = bitcast <vscale x 8 x bfloat>* %base to bfloat *
-%res = call <vscale x 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %Pg, bfloat *%base_ptr)
-ret <vscale x 16 x bfloat> %res
-}
-
-; ld2w
-define <vscale x 8 x i32> @ld2.nxv8i32(<vscale x 4 x i1> %Pg, <vscale x 4 x i32>* %addr) {
-; CHECK-LABEL: ld2.nxv8i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0, #14, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %addr, i64 14
-%base_ptr = bitcast <vscale x 4 x i32>* %base to i32 *
-%res = call <vscale x 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1.p0i32(<vscale x 4 x i1> %Pg, i32 *%base_ptr)
-ret <vscale x 8 x i32> %res
-}
-
-define <vscale x 8 x float> @ld2.nxv8f32(<vscale x 4 x i1> %Pg, <vscale x 4 x float>* %addr) {
-; CHECK-LABEL: ld2.nxv8f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0, #-16, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %addr, i64 -16
-%base_ptr = bitcast <vscale x 4 x float>* %base to float *
-%res = call <vscale x 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1.p0f32(<vscale x 4 x i1> %Pg, float *%base_ptr)
-ret <vscale x 8 x float> %res
-}
-
-; ld2d
-define <vscale x 4 x i64> @ld2.nxv4i64(<vscale x 2 x i1> %Pg, <vscale x 2 x i64>* %addr) {
-; CHECK-LABEL: ld2.nxv4i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0, #14, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %addr, i64 14
-%base_ptr = bitcast <vscale x 2 x i64>* %base to i64 *
-%res = call <vscale x 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1.p0i64(<vscale x 2 x i1> %Pg, i64 *%base_ptr)
-ret <vscale x 4 x i64> %res
-}
-
-define <vscale x 4 x double> @ld2.nxv4f64(<vscale x 2 x i1> %Pg, <vscale x 2 x double>* %addr) {
-; CHECK-LABEL: ld2.nxv4f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0, #-16, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %addr, i64 -16
-%base_ptr = bitcast <vscale x 2 x double>* %base to double *
-%res = call <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1.p0f64(<vscale x 2 x i1> %Pg, double *%base_ptr)
-ret <vscale x 4 x double> %res
-}
-
-; ld3b
-define <vscale x 48 x i8> @ld3.nxv48i8(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, #3, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 3
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-define <vscale x 48 x i8> @ld3.nxv48i8_lower_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8_lower_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, #-24, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 -24
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-define <vscale x 48 x i8> @ld3.nxv48i8_upper_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8_upper_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, #21, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 21
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-define <vscale x 48 x i8> @ld3.nxv48i8_not_multiple_of_3_01(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8_not_multiple_of_3_01:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #4
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 4
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-define <vscale x 48 x i8> @ld3.nxv48i8_not_multiple_of_3_02(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8_not_multiple_of_3_02:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #5
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 5
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-define <vscale x 48 x i8> @ld3.nxv48i8_outside_lower_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8_outside_lower_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #-27
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 -27
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-define <vscale x 48 x i8> @ld3.nxv48i8_outside_upper_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld3.nxv48i8_outside_upper_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #24
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 24
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 48 x i8> %res
-}
-
-; ld3h
-define <vscale x 24 x i16> @ld3.nxv24i16(<vscale x 8 x i1> %Pg, <vscale x 8 x i16> *%addr) {
-; CHECK-LABEL: ld3.nxv24i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0, #21, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %addr, i64 21
-%base_ptr = bitcast <vscale x 8 x i16>* %base to i16 *
-%res = call <vscale x 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1.p0i16(<vscale x 8 x i1> %Pg, i16 *%base_ptr)
-ret <vscale x 24 x i16> %res
-}
-
-define <vscale x 24 x half> @ld3.nxv24f16(<vscale x 8 x i1> %Pg, <vscale x 8 x half> *%addr) {
-; CHECK-LABEL: ld3.nxv24f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0, #21, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %addr, i64 21
-%base_ptr = bitcast <vscale x 8 x half>* %base to half *
-%res = call <vscale x 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1.p0f16(<vscale x 8 x i1> %Pg, half *%base_ptr)
-ret <vscale x 24 x half> %res
-}
-
-define <vscale x 24 x bfloat> @ld3.nxv24bf16(<vscale x 8 x i1> %Pg, <vscale x 8 x bfloat> *%addr) #0 {
-; CHECK-LABEL: ld3.nxv24bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0, #-24, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* %addr, i64 -24
-%base_ptr = bitcast <vscale x 8 x bfloat>* %base to bfloat *
-%res = call <vscale x 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %Pg, bfloat *%base_ptr)
-ret <vscale x 24 x bfloat> %res
-}
-
-; ld3w
-define <vscale x 12 x i32> @ld3.nxv12i32(<vscale x 4 x i1> %Pg, <vscale x 4 x i32> *%addr) {
-; CHECK-LABEL: ld3.nxv12i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3w { z0.s, z1.s, z2.s }, p0/z, [x0, #21, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %addr, i64 21
-%base_ptr = bitcast <vscale x 4 x i32>* %base to i32 *
-%res = call <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1.p0i32(<vscale x 4 x i1> %Pg, i32 *%base_ptr)
-ret <vscale x 12 x i32> %res
-}
-
-define <vscale x 12 x float> @ld3.nxv12f32(<vscale x 4 x i1> %Pg, <vscale x 4 x float> *%addr) {
-; CHECK-LABEL: ld3.nxv12f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3w { z0.s, z1.s, z2.s }, p0/z, [x0, #-24, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %addr, i64 -24
-%base_ptr = bitcast <vscale x 4 x float>* %base to float *
-%res = call <vscale x 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1.p0f32(<vscale x 4 x i1> %Pg, float *%base_ptr)
-ret <vscale x 12 x float> %res
-}
-
-; ld3d
-define <vscale x 6 x i64> @ld3.nxv6i64(<vscale x 2 x i1> %Pg, <vscale x 2 x i64> *%addr) {
-; CHECK-LABEL: ld3.nxv6i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3d { z0.d, z1.d, z2.d }, p0/z, [x0, #21, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %addr, i64 21
-%base_ptr = bitcast <vscale x 2 x i64>* %base to i64 *
-%res = call <vscale x 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1.p0i64(<vscale x 2 x i1> %Pg, i64 *%base_ptr)
-ret <vscale x 6 x i64> %res
-}
-
-define <vscale x 6 x double> @ld3.nxv6f64(<vscale x 2 x i1> %Pg, <vscale x 2 x double> *%addr) {
-; CHECK-LABEL: ld3.nxv6f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3d { z0.d, z1.d, z2.d }, p0/z, [x0, #-24, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %addr, i64 -24
-%base_ptr = bitcast <vscale x 2 x double>* %base to double *
-%res = call <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1.p0f64(<vscale x 2 x i1> %Pg, double *%base_ptr)
-ret <vscale x 6 x double> %res
-}
-
-; ; ld4b
-define <vscale x 64 x i8> @ld4.nxv64i8(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, #4, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 4
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_lower_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_lower_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, #-32, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 -32
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_upper_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_upper_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, #28, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 28
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_not_multiple_of_4_01(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_not_multiple_of_4_01:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #5
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 5
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_not_multiple_of_4_02(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_not_multiple_of_4_02:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #6
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 6
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_not_multiple_of_4_03(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_not_multiple_of_4_03:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #7
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 7
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_outside_lower_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_outside_lower_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #1
-; CHECK-NEXT:    mov x9, #-576
-; CHECK-NEXT:    lsr x8, x8, #4
-; CHECK-NEXT:    mul x8, x8, x9
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-; FIXME: optimize OFFSET computation so that xOFFSET = (mul (RDVL #4) #9)
-; xM = -9 * 2^6
-; xP = RDVL * 2^-4
-; xOFFSET = RDVL * 2^-4 * -9 * 2^6 = RDVL * -36
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 -36
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-define <vscale x 64 x i8> @ld4.nxv64i8_outside_upper_bound(<vscale x 16 x i1> %Pg, <vscale x 16 x i8> *%addr) {
-; CHECK-LABEL: ld4.nxv64i8_outside_upper_bound:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    rdvl x8, #1
-; CHECK-NEXT:    mov w9, #512
-; CHECK-NEXT:    lsr x8, x8, #4
-; CHECK-NEXT:    mul x8, x8, x9
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, x8]
-; CHECK-NEXT:    ret
-; FIXME: optimize OFFSET computation so that xOFFSET = (mul (RDVL #16) #2)
-; xM = 2^9
-; xP = RDVL * 2^-4
-; xOFFSET = RDVL * 2^-4 * 2^9 = RDVL * 32
-%base = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %addr, i64 32
-%base_ptr = bitcast <vscale x 16 x i8>* %base to i8 *
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%base_ptr)
-ret <vscale x 64 x i8> %res
-}
-
-; ld4h
-define <vscale x 32 x i16> @ld4.nxv32i16(<vscale x 8 x i1> %Pg, <vscale x 8 x i16> *%addr) {
-; CHECK-LABEL: ld4.nxv32i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0, #8, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %addr, i64 8
-%base_ptr = bitcast <vscale x 8 x i16>* %base to i16 *
-%res = call <vscale x 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1.p0i16(<vscale x 8 x i1> %Pg, i16 *%base_ptr)
-ret <vscale x 32 x i16> %res
-}
-
-define <vscale x 32 x half> @ld4.nxv32f16(<vscale x 8 x i1> %Pg, <vscale x 8 x half> *%addr) {
-; CHECK-LABEL: ld4.nxv32f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0, #28, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %addr, i64 28
-%base_ptr = bitcast <vscale x 8 x half>* %base to half *
-%res = call <vscale x 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1.p0f16(<vscale x 8 x i1> %Pg, half *%base_ptr)
-ret <vscale x 32 x half> %res
-}
-
-define <vscale x 32 x bfloat> @ld4.nxv32bf16(<vscale x 8 x i1> %Pg, <vscale x 8 x bfloat> *%addr) #0 {
-; CHECK-LABEL: ld4.nxv32bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0, #-32, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 8 x bfloat>, <vscale x 8 x bfloat>* %addr, i64 -32
-%base_ptr = bitcast <vscale x 8 x bfloat>* %base to bfloat *
-%res = call <vscale x 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %Pg, bfloat *%base_ptr)
-ret <vscale x 32 x bfloat> %res
-}
-
-; ld4w
-define <vscale x 16 x i32> @ld4.nxv16i32(<vscale x 4 x i1> %Pg, <vscale x 4 x i32> *%addr) {
-; CHECK-LABEL: ld4.nxv16i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4w { z0.s, z1.s, z2.s, z3.s }, p0/z, [x0, #28, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %addr, i64 28
-%base_ptr = bitcast <vscale x 4 x i32>* %base to i32 *
-%res = call <vscale x 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1.p0i32(<vscale x 4 x i1> %Pg, i32 *%base_ptr)
-ret <vscale x 16 x i32> %res
-}
-
-define <vscale x 16 x float> @ld4.nxv16f32(<vscale x 4 x i1> %Pg, <vscale x 4 x float>* %addr) {
-; CHECK-LABEL: ld4.nxv16f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4w { z0.s, z1.s, z2.s, z3.s }, p0/z, [x0, #-32, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %addr, i64 -32
-%base_ptr = bitcast <vscale x 4 x float>* %base to float *
-%res = call <vscale x 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1.p0f32(<vscale x 4 x i1> %Pg, float *%base_ptr)
-ret <vscale x 16 x float> %res
-}
-
-; ld4d
-define <vscale x 8 x i64> @ld4.nxv8i64(<vscale x 2 x i1> %Pg, <vscale x 2 x i64> *%addr) {
-; CHECK-LABEL: ld4.nxv8i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4d { z0.d, z1.d, z2.d, z3.d }, p0/z, [x0, #28, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %addr, i64 28
-%base_ptr = bitcast <vscale x 2 x i64>* %base to i64 *
-%res = call <vscale x 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1.p0i64(<vscale x 2 x i1> %Pg, i64 *%base_ptr)
-ret <vscale x 8 x i64> %res
-}
-
-define <vscale x 8 x double> @ld4.nxv8f64(<vscale x 2 x i1> %Pg, <vscale x 2 x double> *%addr) {
-; CHECK-LABEL: ld4.nxv8f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4d { z0.d, z1.d, z2.d, z3.d }, p0/z, [x0, #-32, mul vl]
-; CHECK-NEXT:    ret
-%base = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %addr, i64 -32
-%base_ptr = bitcast <vscale x 2 x double>* %base to double *
-%res = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1.p0f64(<vscale x 2 x i1> %Pg, double * %base_ptr)
-ret <vscale x 8 x double> %res
-}
-
-declare <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-declare <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-declare <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-; +bf16 is required for the bfloat version.
-attributes #0 = { "target-features"="+bf16" }
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-ldN-reg+reg-addr-mode.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-ldN-reg+reg-addr-mode.ll
deleted file mode 100644
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-ldN-reg+reg-addr-mode.ll
+++ /dev/null
@@ -1,285 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=aarch64--linux-gnu -mattr=sve < %s | FileCheck %s
-; RUN: llc -mtriple=aarch64--linux-gnu -mattr=sme < %s | FileCheck %s
-
-; ld2b
-define <vscale x 32 x i8> @ld2.nxv32i8(<vscale x 16 x i1> %Pg, i8 *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv32i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0, x1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i8, i8 *  %addr, i64 %a
-%res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%addr2)
-ret <vscale x 32 x i8> %res
-}
-
-; ld2h
-define <vscale x 16 x i16> @ld2.nxv16i16(<vscale x 8 x i1> %Pg, i16 *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv16i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i16, i16 *  %addr, i64 %a
-%res = call <vscale x 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1.p0i16(<vscale x 8 x i1> %Pg, i16 *%addr2)
-ret <vscale x 16 x i16> %res
-}
-
-define <vscale x 16 x half> @ld2.nxv16f16(<vscale x 8 x i1> %Pg, half *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv16f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr half, half *  %addr, i64 %a
-%res = call <vscale x 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1.p0f16(<vscale x 8 x i1> %Pg, half *%addr2)
-ret <vscale x 16 x half> %res
-}
-
-define <vscale x 16 x bfloat> @ld2.nxv16bf16(<vscale x 8 x i1> %Pg, bfloat *%addr, i64 %a) #0 {
-; CHECK-LABEL: ld2.nxv16bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr bfloat, bfloat *  %addr, i64 %a
-%res = call <vscale x 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %Pg, bfloat *%addr2)
-ret <vscale x 16 x bfloat> %res
-}
-
-; ld2w
-define <vscale x 8 x i32> @ld2.nxv8i32(<vscale x 4 x i1> %Pg, i32 *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv8i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i32, i32 *  %addr, i64 %a
-%res = call <vscale x 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1.p0i32(<vscale x 4 x i1> %Pg, i32 *%addr2)
-ret <vscale x 8 x i32> %res
-}
-
-define <vscale x 8 x float> @ld2.nxv8f32(<vscale x 4 x i1> %Pg, float *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv8f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr float, float *  %addr, i64 %a
-%res = call <vscale x 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1.p0f32(<vscale x 4 x i1> %Pg, float *%addr2)
-ret <vscale x 8 x float> %res
-}
-
-; ld2d
-define <vscale x 4 x i64> @ld2.nxv4i64(<vscale x 2 x i1> %Pg, i64 *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv4i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0, x1, lsl #3]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i64, i64 *  %addr, i64 %a
-%res = call <vscale x 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1.p0i64(<vscale x 2 x i1> %Pg, i64 *%addr2)
-ret <vscale x 4 x i64> %res
-}
-
-define <vscale x 4 x double> @ld2.nxv4f64(<vscale x 2 x i1> %Pg, double *%addr, i64 %a) {
-; CHECK-LABEL: ld2.nxv4f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0, x1, lsl #3]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr double, double *  %addr, i64 %a
-%res = call <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1.p0f64(<vscale x 2 x i1> %Pg, double *%addr2)
-ret <vscale x 4 x double> %res
-}
-
-; ld3b
-define <vscale x 48 x i8> @ld3.nxv48i8(<vscale x 16 x i1> %Pg, i8 *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv48i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0, x1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i8, i8 *  %addr, i64 %a
-%res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%addr2)
-ret <vscale x 48 x i8> %res
-}
-
-; ld3h
-define <vscale x 24 x i16> @ld3.nxv24i16(<vscale x 8 x i1> %Pg, i16 *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv24i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i16, i16 *  %addr, i64 %a
-%res = call <vscale x 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1.p0i16(<vscale x 8 x i1> %Pg, i16 *%addr2)
-ret <vscale x 24 x i16> %res
-}
-
-define <vscale x 24 x half> @ld3.nxv24f16(<vscale x 8 x i1> %Pg, half *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv24f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr half, half *  %addr, i64 %a
-%res = call <vscale x 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1.p0f16(<vscale x 8 x i1> %Pg, half *%addr2)
-ret <vscale x 24 x half> %res
-}
-
-define <vscale x 24 x bfloat> @ld3.nxv24bf16(<vscale x 8 x i1> %Pg, bfloat *%addr, i64 %a) #0 {
-; CHECK-LABEL: ld3.nxv24bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr bfloat, bfloat *  %addr, i64 %a
-%res = call <vscale x 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %Pg, bfloat *%addr2)
-ret <vscale x 24 x bfloat> %res
-}
-
-; ld3w
-define <vscale x 12 x i32> @ld3.nxv12i32(<vscale x 4 x i1> %Pg, i32 *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv12i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3w { z0.s, z1.s, z2.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i32, i32 *  %addr, i64 %a
-%res = call <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1.p0i32(<vscale x 4 x i1> %Pg, i32 *%addr2)
-ret <vscale x 12 x i32> %res
-}
-
-define <vscale x 12 x float> @ld3.nxv12f32(<vscale x 4 x i1> %Pg, float *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv12f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3w { z0.s, z1.s, z2.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr float, float *  %addr, i64 %a
-%res = call <vscale x 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1.p0f32(<vscale x 4 x i1> %Pg, float *%addr2)
-ret <vscale x 12 x float> %res
-}
-
-; ld3d
-define <vscale x 6 x i64> @ld3.nxv6i64(<vscale x 2 x i1> %Pg, i64 *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv6i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3d { z0.d, z1.d, z2.d }, p0/z, [x0, x1, lsl #3]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i64, i64 *  %addr, i64 %a
-%res = call <vscale x 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1.p0i64(<vscale x 2 x i1> %Pg, i64 *%addr2)
-ret <vscale x 6 x i64> %res
-}
-
-define <vscale x 6 x double> @ld3.nxv6f64(<vscale x 2 x i1> %Pg, double *%addr, i64 %a) {
-; CHECK-LABEL: ld3.nxv6f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3d { z0.d, z1.d, z2.d }, p0/z, [x0, x1, lsl #3]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr double, double *  %addr, i64 %a
-%res = call <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1.p0f64(<vscale x 2 x i1> %Pg, double *%addr2)
-ret <vscale x 6 x double> %res
-}
-
-; ld4b
-define <vscale x 64 x i8> @ld4.nxv64i8(<vscale x 16 x i1> %Pg, i8 *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv64i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0, x1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i8, i8 *  %addr, i64 %a
-%res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %Pg, i8 *%addr2)
-ret <vscale x 64 x i8> %res
-}
-
-; ld4h
-define <vscale x 32 x i16> @ld4.nxv32i16(<vscale x 8 x i1> %Pg, i16 *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv32i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i16, i16 *  %addr, i64 %a
-%res = call <vscale x 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1.p0i16(<vscale x 8 x i1> %Pg, i16 *%addr2)
-ret <vscale x 32 x i16> %res
-}
-
-define <vscale x 32 x half> @ld4.nxv32f16(<vscale x 8 x i1> %Pg, half *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv32f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr half, half *  %addr, i64 %a
-%res = call <vscale x 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1.p0f16(<vscale x 8 x i1> %Pg, half *%addr2)
-ret <vscale x 32 x half> %res
-}
-
-define <vscale x 32 x bfloat> @ld4.nxv32bf16(<vscale x 8 x i1> %Pg, bfloat *%addr, i64 %a) #0 {
-; CHECK-LABEL: ld4.nxv32bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0, x1, lsl #1]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr bfloat, bfloat *  %addr, i64 %a
-%res = call <vscale x 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %Pg, bfloat *%addr2)
-ret <vscale x 32 x bfloat> %res
-}
-
-; ld4w
-define <vscale x 16 x i32> @ld4.nxv16i32(<vscale x 4 x i1> %Pg, i32 *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv16i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4w { z0.s, z1.s, z2.s, z3.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i32, i32 *  %addr, i64 %a
-%res = call <vscale x 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1.p0i32(<vscale x 4 x i1> %Pg, i32 *%addr2)
-ret <vscale x 16 x i32> %res
-}
-
-define <vscale x 16 x float> @ld4.nxv16f32(<vscale x 4 x i1> %Pg, float *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv16f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4w { z0.s, z1.s, z2.s, z3.s }, p0/z, [x0, x1, lsl #2]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr float, float *  %addr, i64 %a
-%res = call <vscale x 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1.p0f32(<vscale x 4 x i1> %Pg, float *%addr2)
-ret <vscale x 16 x float> %res
-}
-
-; ld4d
-define <vscale x 8 x i64> @ld4.nxv8i64(<vscale x 2 x i1> %Pg, i64 *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv8i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4d { z0.d, z1.d, z2.d, z3.d }, p0/z, [x0, x1, lsl #3]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr i64, i64 *  %addr, i64 %a
-%res = call <vscale x 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1.p0i64(<vscale x 2 x i1> %Pg, i64 *%addr2)
-ret <vscale x 8 x i64> %res
-}
-
-define <vscale x 8 x double> @ld4.nxv8f64(<vscale x 2 x i1> %Pg, double *%addr, i64 %a) {
-; CHECK-LABEL: ld4.nxv8f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4d { z0.d, z1.d, z2.d, z3.d }, p0/z, [x0, x1, lsl #3]
-; CHECK-NEXT:    ret
-%addr2 = getelementptr double, double *  %addr, i64 %a
-%res = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1.p0f64(<vscale x 2 x i1> %Pg, double *%addr2)
-ret <vscale x 8 x double> %res
-}
-
-declare <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-declare <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-declare <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-; +bf16 is required for the bfloat version.
-attributes #0 = { "target-features"="+bf16" }
diff --git a/llvm/test/CodeGen/AArch64/sve-intrinsics-loads.ll b/llvm/test/CodeGen/AArch64/sve-intrinsics-loads.ll
--- a/llvm/test/CodeGen/AArch64/sve-intrinsics-loads.ll
+++ b/llvm/test/CodeGen/AArch64/sve-intrinsics-loads.ll
@@ -307,270 +307,6 @@
   ret <vscale x 2 x double> %res
 }
 
-;
-; LD2B
-;
-
-define <vscale x 32 x i8> @ld2b_i8(<vscale x 16 x i1> %pred, i8* %addr) {
-; CHECK-LABEL: ld2b_i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2b { z0.b, z1.b }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1> %pred, i8* %addr)
-  ret <vscale x 32 x i8> %res
-}
-
-;
-; LD2H
-;
-
-define <vscale x 16 x i16> @ld2h_i16(<vscale x 8 x i1> %pred, i16* %addr) {
-; CHECK-LABEL: ld2h_i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1.p0i16(<vscale x 8 x i1> %pred, i16* %addr)
-  ret <vscale x 16 x i16> %res
-}
-
-define <vscale x 16 x half> @ld2h_f16(<vscale x 8 x i1> %pred, half* %addr) {
-; CHECK-LABEL: ld2h_f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1.p0f16(<vscale x 8 x i1> %pred, half* %addr)
-  ret <vscale x 16 x half> %res
-}
-
-define <vscale x 16 x bfloat> @ld2h_bf16(<vscale x 8 x i1> %pred, bfloat* %addr) {
-; CHECK-LABEL: ld2h_bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2h { z0.h, z1.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %pred, bfloat* %addr)
-  ret <vscale x 16 x bfloat> %res
-}
-
-;
-; LD2W
-;
-
-define <vscale x 8 x i32> @ld2w_i32(<vscale x 4 x i1> %pred, i32* %addr) {
-; CHECK-LABEL: ld2w_i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1.p0i32(<vscale x 4 x i1> %pred, i32* %addr)
-  ret <vscale x 8 x i32> %res
-}
-
-define <vscale x 8 x float> @ld2w_f32(<vscale x 4 x i1> %pred, float* %addr) {
-; CHECK-LABEL: ld2w_f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2w { z0.s, z1.s }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1.p0f32(<vscale x 4 x i1> %pred, float* %addr)
-  ret <vscale x 8 x float> %res
-}
-
-;
-; LD2D
-;
-
-define <vscale x 4 x i64> @ld2d_i64(<vscale x 2 x i1> %pred, i64* %addr) {
-; CHECK-LABEL: ld2d_i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1.p0i64(<vscale x 2 x i1> %pred, i64* %addr)
-  ret <vscale x 4 x i64> %res
-}
-
-define <vscale x 4 x double> @ld2d_f64(<vscale x 2 x i1> %pred, double* %addr) {
-; CHECK-LABEL: ld2d_f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1.p0f64(<vscale x 2 x i1> %pred, double* %addr)
-  ret <vscale x 4 x double> %res
-}
-
-;
-; LD3B
-;
-
-define <vscale x 48 x i8> @ld3b_i8(<vscale x 16 x i1> %pred, i8* %addr) {
-; CHECK-LABEL: ld3b_i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3b { z0.b, z1.b, z2.b }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1> %pred, i8* %addr)
-  ret <vscale x 48 x i8> %res
-}
-
-;
-; LD3H
-;
-
-define <vscale x 24 x i16> @ld3h_i16(<vscale x 8 x i1> %pred, i16* %addr) {
-; CHECK-LABEL: ld3h_i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1.p0i16(<vscale x 8 x i1> %pred, i16* %addr)
-  ret <vscale x 24 x i16> %res
-}
-
-define <vscale x 24 x half> @ld3h_f16(<vscale x 8 x i1> %pred, half* %addr) {
-; CHECK-LABEL: ld3h_f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1.p0f16(<vscale x 8 x i1> %pred, half* %addr)
-  ret <vscale x 24 x half> %res
-}
-
-define <vscale x 24 x bfloat> @ld3h_bf16(<vscale x 8 x i1> %pred, bfloat* %addr) {
-; CHECK-LABEL: ld3h_bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3h { z0.h, z1.h, z2.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %pred, bfloat* %addr)
-  ret <vscale x 24 x bfloat> %res
-}
-
-;
-; LD3W
-;
-
-define <vscale x 12 x i32> @ld3w_i32(<vscale x 4 x i1> %pred, i32* %addr) {
-; CHECK-LABEL: ld3w_i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3w { z0.s, z1.s, z2.s }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1.p0i32(<vscale x 4 x i1> %pred, i32* %addr)
-  ret <vscale x 12 x i32> %res
-}
-
-define <vscale x 12 x float> @ld3w_f32(<vscale x 4 x i1> %pred, float* %addr) {
-; CHECK-LABEL: ld3w_f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3w { z0.s, z1.s, z2.s }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1.p0f32(<vscale x 4 x i1> %pred, float* %addr)
-  ret <vscale x 12 x float> %res
-}
-
-;
-; LD3D
-;
-
-define <vscale x 6 x i64> @ld3d_i64(<vscale x 2 x i1> %pred, i64* %addr) {
-; CHECK-LABEL: ld3d_i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3d { z0.d, z1.d, z2.d }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1.p0i64(<vscale x 2 x i1> %pred, i64* %addr)
-  ret <vscale x 6 x i64> %res
-}
-
-define <vscale x 6 x double> @ld3d_f64(<vscale x 2 x i1> %pred, double* %addr) {
-; CHECK-LABEL: ld3d_f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld3d { z0.d, z1.d, z2.d }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1.p0f64(<vscale x 2 x i1> %pred, double* %addr)
-  ret <vscale x 6 x double> %res
-}
-
-;
-; LD4B
-;
-
-define <vscale x 64 x i8> @ld4b_i8(<vscale x 16 x i1> %pred, i8* %addr) {
-; CHECK-LABEL: ld4b_i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4b { z0.b, z1.b, z2.b, z3.b }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1> %pred, i8* %addr)
-  ret <vscale x 64 x i8> %res
-}
-
-;
-; LD4H
-;
-
-define <vscale x 32 x i16> @ld4h_i16(<vscale x 8 x i1> %pred, i16* %addr) {
-; CHECK-LABEL: ld4h_i16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1.p0i16(<vscale x 8 x i1> %pred, i16* %addr)
-  ret <vscale x 32 x i16> %res
-}
-
-define <vscale x 32 x half> @ld4h_f16(<vscale x 8 x i1> %pred, half* %addr) {
-; CHECK-LABEL: ld4h_f16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1.p0f16(<vscale x 8 x i1> %pred, half* %addr)
-  ret <vscale x 32 x half> %res
-}
-
-define <vscale x 32 x bfloat> @ld4h_bf16(<vscale x 8 x i1> %pred, bfloat* %addr) {
-; CHECK-LABEL: ld4h_bf16:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4h { z0.h, z1.h, z2.h, z3.h }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1.p0bf16(<vscale x 8 x i1> %pred, bfloat* %addr)
-  ret <vscale x 32 x bfloat> %res
-}
-
-;
-; LD4W
-;
-
-define <vscale x 16 x i32> @ld4w_i32(<vscale x 4 x i1> %pred, i32* %addr) {
-; CHECK-LABEL: ld4w_i32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4w { z0.s, z1.s, z2.s, z3.s }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1.p0i32(<vscale x 4 x i1> %pred, i32* %addr)
-  ret <vscale x 16 x i32> %res
-}
-
-define <vscale x 16 x float> @ld4w_f32(<vscale x 4 x i1> %pred, float* %addr) {
-; CHECK-LABEL: ld4w_f32:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4w { z0.s, z1.s, z2.s, z3.s }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1.p0f32(<vscale x 4 x i1> %pred, float* %addr)
-  ret <vscale x 16 x float> %res
-}
-
-;
-; LD4D
-;
-
-define <vscale x 8 x i64> @ld4d_i64(<vscale x 2 x i1> %pred, i64* %addr) {
-; CHECK-LABEL: ld4d_i64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4d { z0.d, z1.d, z2.d, z3.d }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1.p0i64(<vscale x 2 x i1> %pred, i64* %addr)
-  ret <vscale x 8 x i64> %res
-}
-
-define <vscale x 8 x double> @ld4d_f64(<vscale x 2 x i1> %pred, double* %addr) {
-; CHECK-LABEL: ld4d_f64:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld4d { z0.d, z1.d, z2.d, z3.d }, p0/z, [x0]
-; CHECK-NEXT:    ret
-  %res = call <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1.p0f64(<vscale x 2 x i1> %pred, double* %addr)
-  ret <vscale x 8 x double> %res
-}
-
 
 declare <vscale x 16 x i8> @llvm.aarch64.sve.ld1rq.nxv16i8(<vscale x 16 x i1>, i8*)
 declare <vscale x 8 x i16> @llvm.aarch64.sve.ld1rq.nxv8i16(<vscale x 8 x i1>, i16*)
@@ -590,29 +326,3 @@
 declare <vscale x 4 x float> @llvm.aarch64.sve.ldnt1.nxv4f32(<vscale x 4 x i1>, float*)
 declare <vscale x 2 x double> @llvm.aarch64.sve.ldnt1.nxv2f64(<vscale x 2 x i1>, double*)
 
-declare <vscale x 32 x i8> @llvm.aarch64.sve.ld2.nxv32i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 16 x i16> @llvm.aarch64.sve.ld2.nxv16i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 8 x i32> @llvm.aarch64.sve.ld2.nxv8i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 4 x i64> @llvm.aarch64.sve.ld2.nxv4i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 16 x half> @llvm.aarch64.sve.ld2.nxv16f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 16 x bfloat> @llvm.aarch64.sve.ld2.nxv16bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 8 x float> @llvm.aarch64.sve.ld2.nxv8f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-declare <vscale x 48 x i8> @llvm.aarch64.sve.ld3.nxv48i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 24 x i16> @llvm.aarch64.sve.ld3.nxv24i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 12 x i32> @llvm.aarch64.sve.ld3.nxv12i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 6 x i64> @llvm.aarch64.sve.ld3.nxv6i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 24 x half> @llvm.aarch64.sve.ld3.nxv24f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 24 x bfloat> @llvm.aarch64.sve.ld3.nxv24bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 12 x float> @llvm.aarch64.sve.ld3.nxv12f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 6 x double> @llvm.aarch64.sve.ld3.nxv6f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
-
-declare <vscale x 64 x i8> @llvm.aarch64.sve.ld4.nxv64i8.nxv16i1.p0i8(<vscale x 16 x i1>, i8*)
-declare <vscale x 32 x i16> @llvm.aarch64.sve.ld4.nxv32i16.nxv8i1.p0i16(<vscale x 8 x i1>, i16*)
-declare <vscale x 16 x i32> @llvm.aarch64.sve.ld4.nxv16i32.nxv4i1.p0i32(<vscale x 4 x i1>, i32*)
-declare <vscale x 8 x i64> @llvm.aarch64.sve.ld4.nxv8i64.nxv2i1.p0i64(<vscale x 2 x i1>, i64*)
-declare <vscale x 32 x half> @llvm.aarch64.sve.ld4.nxv32f16.nxv8i1.p0f16(<vscale x 8 x i1>, half*)
-declare <vscale x 32 x bfloat> @llvm.aarch64.sve.ld4.nxv32bf16.nxv8i1.p0bf16(<vscale x 8 x i1>, bfloat*)
-declare <vscale x 16 x float> @llvm.aarch64.sve.ld4.nxv16f32.nxv4i1.p0f32(<vscale x 4 x i1>, float*)
-declare <vscale x 8 x double> @llvm.aarch64.sve.ld4.nxv8f64.nxv2i1.p0f64(<vscale x 2 x i1>, double*)
diff --git a/llvm/test/CodeGen/AArch64/sve-merging-stores.ll b/llvm/test/CodeGen/AArch64/sve-merging-stores.ll
--- a/llvm/test/CodeGen/AArch64/sve-merging-stores.ll
+++ b/llvm/test/CodeGen/AArch64/sve-merging-stores.ll
@@ -1,29 +1,30 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
 
 %complex = type { { double, double } }
 
-; Function Attrs: argmemonly nounwind readonly
-declare <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double>, i64 immarg) #3
 
 ; Function Attrs: argmemonly nounwind readonly
-declare <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1(<vscale x 2 x i1>, double*) #3
+declare { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1>, double*) #3
 
 ; Function Attrs: nounwind readnone
 declare double @llvm.aarch64.sve.faddv.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>) #2
 
 define void @foo1(%complex* %outval, <vscale x 2 x i1> %pred, double *%inptr) {
 ; CHECK-LABEL: foo1:
-; CHECK: ld2d { z0.d, z1.d }, p0/z, [x1]
-; CHECK-NEXT: faddv d2, p0, z0.d
-; CHECK-NEXT: faddv d0, p0, z1.d
-; CHECK-NEXT: mov v2.d[1], v0.d[0]
-; CHECK-NEXT: str q2, [x0]
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    ld2d { z0.d, z1.d }, p0/z, [x1]
+; CHECK-NEXT:    faddv d2, p0, z0.d
+; CHECK-NEXT:    faddv d0, p0, z1.d
+; CHECK-NEXT:    mov v2.d[1], v0.d[0]
+; CHECK-NEXT:    str q2, [x0]
+; CHECK-NEXT:    ret
   %realp = getelementptr inbounds %complex, %complex* %outval, i64 0, i32 0, i32 0
   %imagp = getelementptr inbounds %complex, %complex* %outval, i64 0, i32 0, i32 1
-  %1 = call <vscale x 4 x double> @llvm.aarch64.sve.ld2.nxv4f64.nxv2i1(<vscale x 2 x i1> %pred, double* nonnull %inptr)
-  %2 = call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> %1, i64 0)
+  %1 = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2.sret.nxv2f64(<vscale x 2 x i1> %pred, double* nonnull %inptr)
+  %2 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } %1,  0
   %3 = call double @llvm.aarch64.sve.faddv.nxv2f64(<vscale x 2 x i1> %pred, <vscale x 2 x double> %2)
-  %4 = call <vscale x 2 x double> @llvm.vector.extract.nxv2f64.nxv4f64(<vscale x 4 x double> %1, i64 2)
+  %4 = extractvalue { <vscale x 2 x double>, <vscale x 2 x double> } %1, 1
   %5 = call double @llvm.aarch64.sve.faddv.nxv2f64(<vscale x 2 x i1> %pred, <vscale x 2 x double> %4)
   store double %3, double* %realp, align 8
   store double %5, double* %imagp, align 8