[NEON] Support VST1xN intrinsics in AArch32 mode (Clang part)

[kosarev]

Summary

We currently support them only in AArch64. The NEON Reference, however, says they are 'ARMv7, ARMv8' intrinsics.

[kosarev]

Ping.

[SjoerdMeijer]

Agreed: supported architectures are v7/A32/A64.

cfe/trunk/include/clang/Basic/arm_neon.td

 def VLD1_X4   : WInst<"vld1_x4", "4c",
                       "cfhilsUcUiUlUsQcQfQhQiQlQsQUcQUiQUlQUsPcPsQPcQPs">;
 def VLD1_LANE : WInst<"vld1_lane", "dcdi",
                       "QUcQUsQUiQUlQcQsQiQlQhQfQPcQPsUcUsUiUlcsilhfPcPs">;
 def VLD1_DUP  : WInst<"vld1_dup", "dc",
                       "QUcQUsQUiQUlQcQsQiQlQhQfQPcQPsUcUsUiUlcsilhfPcPs">;
 def VST1      : WInst<"vst1", "vpd",
                       "QUcQUsQUiQUlQcQsQiQlQhQfQPcQPsUcUsUiUlcsilhfPcPs">;
+def VST1_X2   : WInst<"vst1_x2", "vp2",
+                      "cfhilsUcUiUlUsQcQfQhQiQlQsQUcQUiQUlQUsPcPsQPcQPs">;
+def VST1_X3   : WInst<"vst1_x3", "vp3",
+                      "cfhilsUcUiUlUsQcQfQhQiQlQsQUcQUiQUlQUsPcPsQPcQPs">;
+def VST1_X4   : WInst<"vst1_x4", "vp4",
+                      "cfhilsUcUiUlUsQcQfQhQiQlQsQUcQUiQUlQUsPcPsQPcQPs">;
 def VST1_LANE : WInst<"vst1_lane", "vpdi",
                       "QUcQUsQUiQUlQcQsQiQlQhQfQPcQPsUcUsUiUlcsilhfPcPs">;
 
 ////////////////////////////////////////////////////////////////////////////////
 // E.3.15 Loads and stores of an N-element structure
 def VLD2 : WInst<"vld2", "2c", "QUcQUsQUiQcQsQiQhQfQPcQPsUcUsUiUlcsilhfPcPs">;
 def VLD3 : WInst<"vld3", "3c", "QUcQUsQUiQcQsQiQhQfQPcQPsUcUsUiUlcsilhfPcPs">;
 def VLD4 : WInst<"vld4", "4c", "QUcQUsQUiQcQsQiQhQfQPcQPsUcUsUiUlcsilhfPcPs">;

 
 def LD1_X2 : WInst<"vld1_x2", "2c",
                    "dQdPlQPl">;
 def LD1_X3 : WInst<"vld1_x3", "3c",
                    "dQdPlQPl">;
 def LD1_X4 : WInst<"vld1_x4", "4c",
                    "dQdPlQPl">;
 
-def ST1_X2 : WInst<"vst1_x2", "vp2",
-                   "QUcQUsQUiQcQsQiQhQfQPcQPsUcUsUiUlcsilhfPcPsQUlQldQdPlQPl">;
-def ST1_X3 : WInst<"vst1_x3", "vp3",
-                   "QUcQUsQUiQcQsQiQhQfQPcQPsUcUsUiUlcsilhfPcPsQUlQldQdPlQPl">;
-def ST1_X4 : WInst<"vst1_x4", "vp4",
-                   "QUcQUsQUiQcQsQiQhQfQPcQPsUcUsUiUlcsilhfPcPsQUlQldQdPlQPl">;
+def ST1_X2 : WInst<"vst1_x2", "vp2", "dQdPlQPl">;
+def ST1_X3 : WInst<"vst1_x3", "vp3", "dQdPlQPl">;
+def ST1_X4 : WInst<"vst1_x4", "vp4", "dQdPlQPl">;
 
 def LD1_LANE : WInst<"vld1_lane", "dcdi", "dQdPlQPl">;
 def LD2_LANE : WInst<"vld2_lane", "2c2i", "lUlQcQUcQPcQlQUldQdPlQPl">;
 def LD3_LANE : WInst<"vld3_lane", "3c3i", "lUlQcQUcQPcQlQUldQdPlQPl">;
 def LD4_LANE : WInst<"vld4_lane", "4c4i", "lUlQcQUcQPcQlQUldQdPlQPl">;
 def ST1_LANE : WInst<"vst1_lane", "vpdi", "dQdPlQPl">;
 def ST2_LANE : WInst<"vst2_lane", "vp2i", "lUlQcQUcQPcQlQUldQdPlQPl">;
 def ST3_LANE : WInst<"vst3_lane", "vp3i", "lUlQcQUcQPcQlQUldQdPlQPl">;

cfe/trunk/lib/CodeGen/CGBuiltin.cpp

static const NeonIntrinsicInfo ARMSIMDIntrinsicMap [] = {
   NEONMAP2(vshl_v, arm_neon_vshiftu, arm_neon_vshifts, Add1ArgType | UnsignedAlts),
   NEONMAP0(vshll_n_v),
   NEONMAP0(vshlq_n_v),
   NEONMAP2(vshlq_v, arm_neon_vshiftu, arm_neon_vshifts, Add1ArgType | UnsignedAlts),
   NEONMAP0(vshr_n_v),
   NEONMAP0(vshrn_n_v),
   NEONMAP0(vshrq_n_v),
   NEONMAP1(vst1_v, arm_neon_vst1, 0),
+  NEONMAP1(vst1_x2_v, arm_neon_vst1x2, 0),
+  NEONMAP1(vst1_x3_v, arm_neon_vst1x3, 0),
+  NEONMAP1(vst1_x4_v, arm_neon_vst1x4, 0),
   NEONMAP1(vst1q_v, arm_neon_vst1, 0),
+  NEONMAP1(vst1q_x2_v, arm_neon_vst1x2, 0),
+  NEONMAP1(vst1q_x3_v, arm_neon_vst1x3, 0),
+  NEONMAP1(vst1q_x4_v, arm_neon_vst1x4, 0),
   NEONMAP1(vst2_lane_v, arm_neon_vst2lane, 0),
   NEONMAP1(vst2_v, arm_neon_vst2, 0),
   NEONMAP1(vst2q_lane_v, arm_neon_vst2lane, 0),
   NEONMAP1(vst2q_v, arm_neon_vst2, 0),
   NEONMAP1(vst3_lane_v, arm_neon_vst3lane, 0),
   NEONMAP1(vst3_v, arm_neon_vst3, 0),
   NEONMAP1(vst3q_lane_v, arm_neon_vst3lane, 0),
   NEONMAP1(vst3q_v, arm_neon_vst3, 0),
static const NeonIntrinsicInfo AArch64SIMDIntrinsicMap[] = {
   NEONMAP0(vshl_n_v),
   NEONMAP2(vshl_v, aarch64_neon_ushl, aarch64_neon_sshl, Add1ArgType | UnsignedAlts),
   NEONMAP0(vshll_n_v),
   NEONMAP0(vshlq_n_v),
   NEONMAP2(vshlq_v, aarch64_neon_ushl, aarch64_neon_sshl, Add1ArgType | UnsignedAlts),
   NEONMAP0(vshr_n_v),
   NEONMAP0(vshrn_n_v),
   NEONMAP0(vshrq_n_v),
+  NEONMAP1(vst1_x2_v, aarch64_neon_st1x2, 0),
+  NEONMAP1(vst1_x3_v, aarch64_neon_st1x3, 0),
+  NEONMAP1(vst1_x4_v, aarch64_neon_st1x4, 0),
+  NEONMAP1(vst1q_x2_v, aarch64_neon_st1x2, 0),
+  NEONMAP1(vst1q_x3_v, aarch64_neon_st1x3, 0),
+  NEONMAP1(vst1q_x4_v, aarch64_neon_st1x4, 0),
   NEONMAP0(vsubhn_v),
   NEONMAP0(vtst_v),
   NEONMAP0(vtstq_v),
 };
 
 static const NeonIntrinsicInfo AArch64SISDIntrinsicMap[] = {
   NEONMAP1(vabdd_f64, aarch64_sisd_fabd, Add1ArgType),
   NEONMAP1(vabds_f32, aarch64_sisd_fabd, Add1ArgType),
Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
   case NEON::BI__builtin_neon_vst3_lane_v:
   case NEON::BI__builtin_neon_vst3q_lane_v:
   case NEON::BI__builtin_neon_vst4_lane_v:
   case NEON::BI__builtin_neon_vst4q_lane_v: {
     llvm::Type *Tys[] = {Int8PtrTy, Ty};
     Ops.push_back(getAlignmentValue32(PtrOp0));
     return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "");
   }
+  case NEON::BI__builtin_neon_vst1_x2_v:
+  case NEON::BI__builtin_neon_vst1q_x2_v:
+  case NEON::BI__builtin_neon_vst1_x3_v:
+  case NEON::BI__builtin_neon_vst1q_x3_v:
+  case NEON::BI__builtin_neon_vst1_x4_v:
+  case NEON::BI__builtin_neon_vst1q_x4_v: {
+    llvm::Type *PTy = llvm::PointerType::getUnqual(VTy->getVectorElementType());
+    // TODO: Currently in AArch32 mode the pointer operand comes first, whereas
+    // in AArch64 it comes last. We may want to stick to one or another.
+    if (Arch == llvm::Triple::aarch64 || Arch == llvm::Triple::aarch64_be) {
+      llvm::Type *Tys[2] = { VTy, PTy };
+      std::rotate(Ops.begin(), Ops.begin() + 1, Ops.end());
+      return EmitNeonCall(CGM.getIntrinsic(LLVMIntrinsic, Tys), Ops, "");
+    }
+    llvm::Type *Tys[2] = { PTy, VTy };
+    return EmitNeonCall(CGM.getIntrinsic(LLVMIntrinsic, Tys), Ops, "");
+  }
   case NEON::BI__builtin_neon_vsubhn_v: {
     llvm::VectorType *SrcTy =
         llvm::VectorType::getExtendedElementVectorType(VTy);
 
     // %sum = add <4 x i32> %lhs, %rhs
     Ops[0] = Builder.CreateBitCast(Ops[0], SrcTy);
     Ops[1] = Builder.CreateBitCast(Ops[1], SrcTy);
     Ops[0] = Builder.CreateSub(Ops[0], Ops[1], "vsubhn");
case NEON::BI__builtin_neon_vrsraq_n_v: {
     SmallVector<llvm::Value*,2> TmpOps;
     TmpOps.push_back(Ops[1]);
     TmpOps.push_back(Ops[2]);
     Function* F = CGM.getIntrinsic(Int, Ty);
     llvm::Value *tmp = EmitNeonCall(F, TmpOps, "vrshr_n", 1, true);
     Ops[0] = Builder.CreateBitCast(Ops[0], VTy);
     return Builder.CreateAdd(Ops[0], tmp);
   }
-    // FIXME: Sharing loads & stores with 32-bit is complicated by the absence
-    // of an Align parameter here.
-  case NEON::BI__builtin_neon_vst1_x2_v:
-  case NEON::BI__builtin_neon_vst1q_x2_v:
-  case NEON::BI__builtin_neon_vst1_x3_v:
-  case NEON::BI__builtin_neon_vst1q_x3_v:
-  case NEON::BI__builtin_neon_vst1_x4_v:
-  case NEON::BI__builtin_neon_vst1q_x4_v: {
-    llvm::Type *PTy = llvm::PointerType::getUnqual(VTy->getVectorElementType());
-    llvm::Type *Tys[2] = { VTy, PTy };
-    unsigned Int;
-    switch (BuiltinID) {
-    case NEON::BI__builtin_neon_vst1_x2_v:
-    case NEON::BI__builtin_neon_vst1q_x2_v:
-      Int = Intrinsic::aarch64_neon_st1x2;
-      break;
-    case NEON::BI__builtin_neon_vst1_x3_v:
-    case NEON::BI__builtin_neon_vst1q_x3_v:
-      Int = Intrinsic::aarch64_neon_st1x3;
-      break;
-    case NEON::BI__builtin_neon_vst1_x4_v:
-    case NEON::BI__builtin_neon_vst1q_x4_v:
-      Int = Intrinsic::aarch64_neon_st1x4;
-      break;
-    }
-    std::rotate(Ops.begin(), Ops.begin() + 1, Ops.end());
-    return EmitNeonCall(CGM.getIntrinsic(Int, Tys), Ops, "");
-  }
   case NEON::BI__builtin_neon_vld1_v:
   case NEON::BI__builtin_neon_vld1q_v: {
     Ops[0] = Builder.CreateBitCast(Ops[0], llvm::PointerType::getUnqual(VTy));
     auto Alignment = CharUnits::fromQuantity(
         BuiltinID == NEON::BI__builtin_neon_vld1_v ? 8 : 16);
     return Builder.CreateAlignedLoad(VTy, Ops[0], Alignment);
   }
   case NEON::BI__builtin_neon_vst1_v:

cfe/trunk/test/CodeGen/aarch64-neon-intrinsics.c

 // CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 32, i1 false)
 // CHECK:   [[TMP6:%.*]] = load %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[RETVAL]], align 8
 // CHECK:   ret %struct.poly64x1x4_t [[TMP6]]
 poly64x1x4_t test_vld1_p64_x4(poly64_t const *a) {
   return vld1_p64_x4(a);
 }
 
-// CHECK-LABEL: @test_vst1q_u8_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint8x16x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x16x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <16 x i8>] [[B]].coerce, [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint8x16x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint8x16x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_u8_x2(uint8_t *a, uint8x16x2_t b) {
-  vst1q_u8_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint16x8x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x8x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x i16>] [[B]].coerce, [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint16x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint16x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8i16.p0i16(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i16* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_u16_x2(uint16_t *a, uint16x8x2_t b) {
-  vst1q_u16_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u32_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint32x4x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x4x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x i32>] [[B]].coerce, [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint32x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint32x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4i32.p0i32(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i32* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_u32_x2(uint32_t *a, uint32x4x2_t b) {
-  vst1q_u32_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u64_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint64x2x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x2x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <2 x i64>] [[B]].coerce, [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x2x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint64x2x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v2i64.p0i64(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_u64_x2(uint64_t *a, uint64x2x2_t b) {
-  vst1q_u64_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s8_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int8x16x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int8x16x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <16 x i8>] [[B]].coerce, [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int8x16x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int8x16x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_s8_x2(int8_t *a, int8x16x2_t b) {
-  vst1q_s8_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int16x8x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int16x8x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x i16>] [[B]].coerce, [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int16x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int16x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8i16.p0i16(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i16* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_s16_x2(int16_t *a, int16x8x2_t b) {
-  vst1q_s16_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s32_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int32x4x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int32x4x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x i32>] [[B]].coerce, [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int32x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int32x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4i32.p0i32(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i32* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_s32_x2(int32_t *a, int32x4x2_t b) {
-  vst1q_s32_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s64_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int64x2x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int64x2x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x2x2_t, %struct.int64x2x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <2 x i64>] [[B]].coerce, [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x2x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int64x2x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int64x2x2_t, %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int64x2x2_t, %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v2i64.p0i64(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_s64_x2(int64_t *a, int64x2x2_t b) {
-  vst1q_s64_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_f16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.float16x8x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.float16x8x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x half>] [[B]].coerce, [2 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float16x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float16x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast half* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x half>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x half>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to half*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8f16.p0f16(<8 x half> [[TMP7]], <8 x half> [[TMP8]], half* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_f16_x2(float16_t *a, float16x8x2_t b) {
-  vst1q_f16_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_f32_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.float32x4x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.float32x4x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x float>] [[B]].coerce, [2 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float32x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float32x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast float* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to float*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4f32.p0f32(<4 x float> [[TMP7]], <4 x float> [[TMP8]], float* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_f32_x2(float32_t *a, float32x4x2_t b) {
-  vst1q_f32_x2(a, b);
-}
-
 // CHECK-LABEL: @test_vst1q_f64_x2(
 // CHECK:   [[B:%.*]] = alloca %struct.float64x2x2_t, align 16
 // CHECK:   [[__S1:%.*]] = alloca %struct.float64x2x2_t, align 16
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float64x2x2_t, %struct.float64x2x2_t* [[B]], i32 0, i32 0
 // CHECK:   store [2 x <2 x double>] [[B]].coerce, [2 x <2 x double>]* [[COERCE_DIVE]], align 16
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x2x2_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.float64x2x2_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)

 // CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
 // CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to double*
 // CHECK:   call void @llvm.aarch64.neon.st1x2.v2f64.p0f64(<2 x double> [[TMP7]], <2 x double> [[TMP8]], double* [[TMP9]])
 // CHECK:   ret void
 void test_vst1q_f64_x2(float64_t *a, float64x2x2_t b) {
   vst1q_f64_x2(a, b);
 }
 
-// CHECK-LABEL: @test_vst1q_p8_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.poly8x16x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x16x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <16 x i8>] [[B]].coerce, [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly8x16x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly8x16x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_p8_x2(poly8_t *a, poly8x16x2_t b) {
-  vst1q_p8_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_p16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.poly16x8x2_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x8x2_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x i16>] [[B]].coerce, [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly16x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly16x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8i16.p0i16(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i16* [[TMP9]])
-// CHECK:   ret void
-void test_vst1q_p16_x2(poly16_t *a, poly16x8x2_t b) {
-  vst1q_p16_x2(a, b);
-}
-
 // CHECK-LABEL: @test_vst1q_p64_x2(
 // CHECK:   [[B:%.*]] = alloca %struct.poly64x2x2_t, align 16
 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x2_t, align 16
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[B]], i32 0, i32 0
 // CHECK:   store [2 x <2 x i64>] [[B]].coerce, [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x2x2_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)

 // CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
 // CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
 // CHECK:   call void @llvm.aarch64.neon.st1x2.v2i64.p0i64(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64* [[TMP9]])
 // CHECK:   ret void
 void test_vst1q_p64_x2(poly64_t *a, poly64x2x2_t b) {
   vst1q_p64_x2(a, b);
 }
 
-// CHECK-LABEL: @test_vst1_u8_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint8x8x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x8x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x i8>] [[B]].coerce, [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint8x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint8x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i8* %a)
-// CHECK:   ret void
-void test_vst1_u8_x2(uint8_t *a, uint8x8x2_t b) {
-  vst1_u8_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint16x4x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x4x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x i16>] [[B]].coerce, [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint16x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint16x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4i16.p0i16(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i16* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_u16_x2(uint16_t *a, uint16x4x2_t b) {
-  vst1_u16_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u32_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint32x2x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x2x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <2 x i32>] [[B]].coerce, [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint32x2x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint32x2x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v2i32.p0i32(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i32* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_u32_x2(uint32_t *a, uint32x2x2_t b) {
-  vst1_u32_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u64_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.uint64x1x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x1x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <1 x i64>] [[B]].coerce, [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x1x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint64x1x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v1i64.p0i64(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_u64_x2(uint64_t *a, uint64x1x2_t b) {
-  vst1_u64_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s8_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int8x8x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int8x8x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x i8>] [[B]].coerce, [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int8x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int8x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i8* %a)
-// CHECK:   ret void
-void test_vst1_s8_x2(int8_t *a, int8x8x2_t b) {
-  vst1_s8_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int16x4x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int16x4x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x i16>] [[B]].coerce, [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int16x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int16x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4i16.p0i16(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i16* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_s16_x2(int16_t *a, int16x4x2_t b) {
-  vst1_s16_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s32_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int32x2x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int32x2x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <2 x i32>] [[B]].coerce, [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int32x2x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int32x2x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v2i32.p0i32(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i32* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_s32_x2(int32_t *a, int32x2x2_t b) {
-  vst1_s32_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s64_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.int64x1x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int64x1x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x1x2_t, %struct.int64x1x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <1 x i64>] [[B]].coerce, [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x1x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int64x1x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int64x1x2_t, %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int64x1x2_t, %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v1i64.p0i64(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_s64_x2(int64_t *a, int64x1x2_t b) {
-  vst1_s64_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_f16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.float16x4x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.float16x4x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x half>] [[B]].coerce, [2 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float16x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float16x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast half* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x half>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x half>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to half*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4f16.p0f16(<4 x half> [[TMP7]], <4 x half> [[TMP8]], half* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_f16_x2(float16_t *a, float16x4x2_t b) {
-  vst1_f16_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_f32_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.float32x2x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.float32x2x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <2 x float>] [[B]].coerce, [2 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float32x2x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float32x2x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast float* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to float*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v2f32.p0f32(<2 x float> [[TMP7]], <2 x float> [[TMP8]], float* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_f32_x2(float32_t *a, float32x2x2_t b) {
-  vst1_f32_x2(a, b);
-}
-
 // CHECK-LABEL: @test_vst1_f64_x2(
 // CHECK:   [[B:%.*]] = alloca %struct.float64x1x2_t, align 8
 // CHECK:   [[__S1:%.*]] = alloca %struct.float64x1x2_t, align 8
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float64x1x2_t, %struct.float64x1x2_t* [[B]], i32 0, i32 0
 // CHECK:   store [2 x <1 x double>] [[B]].coerce, [2 x <1 x double>]* [[COERCE_DIVE]], align 8
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x1x2_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.float64x1x2_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)

 // CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
 // CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to double*
 // CHECK:   call void @llvm.aarch64.neon.st1x2.v1f64.p0f64(<1 x double> [[TMP7]], <1 x double> [[TMP8]], double* [[TMP9]])
 // CHECK:   ret void
 void test_vst1_f64_x2(float64_t *a, float64x1x2_t b) {
   vst1_f64_x2(a, b);
 }
 
-// CHECK-LABEL: @test_vst1_p8_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.poly8x8x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x8x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <8 x i8>] [[B]].coerce, [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly8x8x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly8x8x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i8* %a)
-// CHECK:   ret void
-void test_vst1_p8_x2(poly8_t *a, poly8x8x2_t b) {
-  vst1_p8_x2(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_p16_x2(
-// CHECK:   [[B:%.*]] = alloca %struct.poly16x4x2_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x4x2_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[B]], i32 0, i32 0
-// CHECK:   store [2 x <4 x i16>] [[B]].coerce, [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly16x4x2_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly16x4x2_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x2.v4i16.p0i16(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i16* [[TMP9]])
-// CHECK:   ret void
-void test_vst1_p16_x2(poly16_t *a, poly16x4x2_t b) {
-  vst1_p16_x2(a, b);
-}
-
 // CHECK-LABEL: @test_vst1_p64_x2(
 // CHECK:   [[B:%.*]] = alloca %struct.poly64x1x2_t, align 8
 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x2_t, align 8
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[B]], i32 0, i32 0
 // CHECK:   store [2 x <1 x i64>] [[B]].coerce, [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x1x2_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)

 // CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
 // CHECK:   [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
 // CHECK:   call void @llvm.aarch64.neon.st1x2.v1i64.p0i64(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64* [[TMP9]])
 // CHECK:   ret void
 void test_vst1_p64_x2(poly64_t *a, poly64x1x2_t b) {
   vst1_p64_x2(a, b);
 }
 
-// CHECK-LABEL: @test_vst1q_u8_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint8x16x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x16x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <16 x i8>] [[B]].coerce, [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint8x16x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint8x16x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_u8_x3(uint8_t *a, uint8x16x3_t b) {
-  vst1q_u8_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint16x8x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x8x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x i16>] [[B]].coerce, [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint16x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint16x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8i16.p0i16(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i16* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_u16_x3(uint16_t *a, uint16x8x3_t b) {
-  vst1q_u16_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u32_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint32x4x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x4x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x i32>] [[B]].coerce, [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint32x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint32x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4i32.p0i32(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i32* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_u32_x3(uint32_t *a, uint32x4x3_t b) {
-  vst1q_u32_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u64_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint64x2x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x2x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <2 x i64>] [[B]].coerce, [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x2x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint64x2x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v2i64.p0i64(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_u64_x3(uint64_t *a, uint64x2x3_t b) {
-  vst1q_u64_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s8_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int8x16x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int8x16x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <16 x i8>] [[B]].coerce, [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int8x16x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int8x16x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_s8_x3(int8_t *a, int8x16x3_t b) {
-  vst1q_s8_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int16x8x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int16x8x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x i16>] [[B]].coerce, [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int16x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int16x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8i16.p0i16(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i16* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_s16_x3(int16_t *a, int16x8x3_t b) {
-  vst1q_s16_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s32_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int32x4x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int32x4x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x i32>] [[B]].coerce, [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int32x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int32x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4i32.p0i32(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i32* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_s32_x3(int32_t *a, int32x4x3_t b) {
-  vst1q_s32_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s64_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int64x2x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int64x2x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <2 x i64>] [[B]].coerce, [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x2x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int64x2x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v2i64.p0i64(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_s64_x3(int64_t *a, int64x2x3_t b) {
-  vst1q_s64_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_f16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.float16x8x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.float16x8x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x half>] [[B]].coerce, [3 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float16x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float16x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast half* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x half> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x half>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x half>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x half>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to half*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8f16.p0f16(<8 x half> [[TMP9]], <8 x half> [[TMP10]], <8 x half> [[TMP11]], half* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_f16_x3(float16_t *a, float16x8x3_t b) {
-  vst1q_f16_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_f32_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.float32x4x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.float32x4x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x float>] [[B]].coerce, [3 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float32x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float32x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast float* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x float> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x float>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to float*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4f32.p0f32(<4 x float> [[TMP9]], <4 x float> [[TMP10]], <4 x float> [[TMP11]], float* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_f32_x3(float32_t *a, float32x4x3_t b) {
-  vst1q_f32_x3(a, b);
-}
-
 // CHECK-LABEL: @test_vst1q_f64_x3(
 // CHECK:   [[B:%.*]] = alloca %struct.float64x2x3_t, align 16
 // CHECK:   [[__S1:%.*]] = alloca %struct.float64x2x3_t, align 16
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[B]], i32 0, i32 0
 // CHECK:   store [3 x <2 x double>] [[B]].coerce, [3 x <2 x double>]* [[COERCE_DIVE]], align 16
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x2x3_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.float64x2x3_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)

 // CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x double>
 // CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to double*
 // CHECK:   call void @llvm.aarch64.neon.st1x3.v2f64.p0f64(<2 x double> [[TMP9]], <2 x double> [[TMP10]], <2 x double> [[TMP11]], double* [[TMP12]])
 // CHECK:   ret void
 void test_vst1q_f64_x3(float64_t *a, float64x2x3_t b) {
   vst1q_f64_x3(a, b);
 }
 
-// CHECK-LABEL: @test_vst1q_p8_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.poly8x16x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x16x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <16 x i8>] [[B]].coerce, [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly8x16x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly8x16x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_p8_x3(poly8_t *a, poly8x16x3_t b) {
-  vst1q_p8_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_p16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.poly16x8x3_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x8x3_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x i16>] [[B]].coerce, [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly16x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly16x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8i16.p0i16(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i16* [[TMP12]])
-// CHECK:   ret void
-void test_vst1q_p16_x3(poly16_t *a, poly16x8x3_t b) {
-  vst1q_p16_x3(a, b);
-}
-
 // CHECK-LABEL: @test_vst1q_p64_x3(
 // CHECK:   [[B:%.*]] = alloca %struct.poly64x2x3_t, align 16
 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x3_t, align 16
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[B]], i32 0, i32 0
 // CHECK:   store [3 x <2 x i64>] [[B]].coerce, [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x2x3_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)

 // CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
 // CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
 // CHECK:   call void @llvm.aarch64.neon.st1x3.v2i64.p0i64(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64* [[TMP12]])
 // CHECK:   ret void
 void test_vst1q_p64_x3(poly64_t *a, poly64x2x3_t b) {
   vst1q_p64_x3(a, b);
 }
 
-// CHECK-LABEL: @test_vst1_u8_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint8x8x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x8x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x i8>] [[B]].coerce, [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint8x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint8x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i8* %a)
-// CHECK:   ret void
-void test_vst1_u8_x3(uint8_t *a, uint8x8x3_t b) {
-  vst1_u8_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint16x4x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x4x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x i16>] [[B]].coerce, [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint16x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint16x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4i16.p0i16(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i16* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_u16_x3(uint16_t *a, uint16x4x3_t b) {
-  vst1_u16_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u32_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint32x2x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x2x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <2 x i32>] [[B]].coerce, [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint32x2x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint32x2x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v2i32.p0i32(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i32* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_u32_x3(uint32_t *a, uint32x2x3_t b) {
-  vst1_u32_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u64_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.uint64x1x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x1x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <1 x i64>] [[B]].coerce, [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x1x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint64x1x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v1i64.p0i64(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_u64_x3(uint64_t *a, uint64x1x3_t b) {
-  vst1_u64_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s8_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int8x8x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int8x8x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x i8>] [[B]].coerce, [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int8x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int8x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i8* %a)
-// CHECK:   ret void
-void test_vst1_s8_x3(int8_t *a, int8x8x3_t b) {
-  vst1_s8_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int16x4x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int16x4x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x i16>] [[B]].coerce, [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int16x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int16x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4i16.p0i16(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i16* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_s16_x3(int16_t *a, int16x4x3_t b) {
-  vst1_s16_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s32_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int32x2x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int32x2x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <2 x i32>] [[B]].coerce, [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int32x2x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int32x2x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v2i32.p0i32(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i32* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_s32_x3(int32_t *a, int32x2x3_t b) {
-  vst1_s32_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s64_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.int64x1x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int64x1x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <1 x i64>] [[B]].coerce, [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x1x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int64x1x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v1i64.p0i64(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_s64_x3(int64_t *a, int64x1x3_t b) {
-  vst1_s64_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_f16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.float16x4x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.float16x4x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x half>] [[B]].coerce, [3 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float16x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float16x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast half* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x half> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x half>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x half>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x half>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to half*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4f16.p0f16(<4 x half> [[TMP9]], <4 x half> [[TMP10]], <4 x half> [[TMP11]], half* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_f16_x3(float16_t *a, float16x4x3_t b) {
-  vst1_f16_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_f32_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.float32x2x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.float32x2x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <2 x float>] [[B]].coerce, [3 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float32x2x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float32x2x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast float* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x float> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x float>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to float*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v2f32.p0f32(<2 x float> [[TMP9]], <2 x float> [[TMP10]], <2 x float> [[TMP11]], float* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_f32_x3(float32_t *a, float32x2x3_t b) {
-  vst1_f32_x3(a, b);
-}
-
 // CHECK-LABEL: @test_vst1_f64_x3(
 // CHECK:   [[B:%.*]] = alloca %struct.float64x1x3_t, align 8
 // CHECK:   [[__S1:%.*]] = alloca %struct.float64x1x3_t, align 8
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[B]], i32 0, i32 0
 // CHECK:   store [3 x <1 x double>] [[B]].coerce, [3 x <1 x double>]* [[COERCE_DIVE]], align 8
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x1x3_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.float64x1x3_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)

 // CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x double>
 // CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to double*
 // CHECK:   call void @llvm.aarch64.neon.st1x3.v1f64.p0f64(<1 x double> [[TMP9]], <1 x double> [[TMP10]], <1 x double> [[TMP11]], double* [[TMP12]])
 // CHECK:   ret void
 void test_vst1_f64_x3(float64_t *a, float64x1x3_t b) {
   vst1_f64_x3(a, b);
 }
 
-// CHECK-LABEL: @test_vst1_p8_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.poly8x8x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x8x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <8 x i8>] [[B]].coerce, [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly8x8x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly8x8x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i8* %a)
-// CHECK:   ret void
-void test_vst1_p8_x3(poly8_t *a, poly8x8x3_t b) {
-  vst1_p8_x3(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_p16_x3(
-// CHECK:   [[B:%.*]] = alloca %struct.poly16x4x3_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x4x3_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[B]], i32 0, i32 0
-// CHECK:   store [3 x <4 x i16>] [[B]].coerce, [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly16x4x3_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly16x4x3_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x3.v4i16.p0i16(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i16* [[TMP12]])
-// CHECK:   ret void
-void test_vst1_p16_x3(poly16_t *a, poly16x4x3_t b) {
-  vst1_p16_x3(a, b);
-}
-
 // CHECK-LABEL: @test_vst1_p64_x3(
 // CHECK:   [[B:%.*]] = alloca %struct.poly64x1x3_t, align 8
 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x3_t, align 8
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[B]], i32 0, i32 0
 // CHECK:   store [3 x <1 x i64>] [[B]].coerce, [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x1x3_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)

 // CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
 // CHECK:   [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
 // CHECK:   call void @llvm.aarch64.neon.st1x3.v1i64.p0i64(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64* [[TMP12]])
 // CHECK:   ret void
 void test_vst1_p64_x3(poly64_t *a, poly64x1x3_t b) {
   vst1_p64_x3(a, b);
 }
 
-// CHECK-LABEL: @test_vst1q_u8_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint8x16x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x16x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <16 x i8>] [[B]].coerce, [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint8x16x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint8x16x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_u8_x4(uint8_t *a, uint8x16x4_t b) {
-  vst1q_u8_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint16x8x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x8x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x i16>] [[B]].coerce, [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint16x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint16x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8i16.p0i16(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i16* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_u16_x4(uint16_t *a, uint16x8x4_t b) {
-  vst1q_u16_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u32_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint32x4x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x4x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x i32>] [[B]].coerce, [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint32x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint32x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x i32> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x i32>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4i32.p0i32(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i32* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_u32_x4(uint32_t *a, uint32x4x4_t b) {
-  vst1q_u32_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_u64_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint64x2x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x2x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <2 x i64>] [[B]].coerce, [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x2x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint64x2x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v2i64.p0i64(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_u64_x4(uint64_t *a, uint64x2x4_t b) {
-  vst1q_u64_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s8_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int8x16x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int8x16x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <16 x i8>] [[B]].coerce, [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int8x16x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int8x16x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_s8_x4(int8_t *a, int8x16x4_t b) {
-  vst1q_s8_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int16x8x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int16x8x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x i16>] [[B]].coerce, [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int16x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int16x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8i16.p0i16(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i16* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_s16_x4(int16_t *a, int16x8x4_t b) {
-  vst1q_s16_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s32_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int32x4x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int32x4x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x i32>] [[B]].coerce, [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int32x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int32x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x i32> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x i32>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4i32.p0i32(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i32* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_s32_x4(int32_t *a, int32x4x4_t b) {
-  vst1q_s32_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_s64_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int64x2x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.int64x2x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <2 x i64>] [[B]].coerce, [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x2x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int64x2x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v2i64.p0i64(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_s64_x4(int64_t *a, int64x2x4_t b) {
-  vst1q_s64_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_f16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.float16x8x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.float16x8x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x half>] [[B]].coerce, [4 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float16x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float16x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast half* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x half> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x half> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x half>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x half>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x half>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x half>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to half*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8f16.p0f16(<8 x half> [[TMP11]], <8 x half> [[TMP12]], <8 x half> [[TMP13]], <8 x half> [[TMP14]], half* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_f16_x4(float16_t *a, float16x8x4_t b) {
-  vst1q_f16_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_f32_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.float32x4x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.float32x4x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x float>] [[B]].coerce, [4 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float32x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float32x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast float* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x float> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x float> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x float>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x float>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to float*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4f32.p0f32(<4 x float> [[TMP11]], <4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]], float* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_f32_x4(float32_t *a, float32x4x4_t b) {
-  vst1q_f32_x4(a, b);
-}
-
 // CHECK-LABEL: @test_vst1q_f64_x4(
 // CHECK:   [[B:%.*]] = alloca %struct.float64x2x4_t, align 16
 // CHECK:   [[__S1:%.*]] = alloca %struct.float64x2x4_t, align 16
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[B]], i32 0, i32 0
 // CHECK:   store [4 x <2 x double>] [[B]].coerce, [4 x <2 x double>]* [[COERCE_DIVE]], align 16
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x2x4_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.float64x2x4_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)

 // CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x double>
 // CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to double*
 // CHECK:   call void @llvm.aarch64.neon.st1x4.v2f64.p0f64(<2 x double> [[TMP11]], <2 x double> [[TMP12]], <2 x double> [[TMP13]], <2 x double> [[TMP14]], double* [[TMP15]])
 // CHECK:   ret void
 void test_vst1q_f64_x4(float64_t *a, float64x2x4_t b) {
   vst1q_f64_x4(a, b);
 }
 
-// CHECK-LABEL: @test_vst1q_p8_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.poly8x16x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x16x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <16 x i8>] [[B]].coerce, [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly8x16x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly8x16x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i8* %a)
-// CHECK:   ret void
-void test_vst1q_p8_x4(poly8_t *a, poly8x16x4_t b) {
-  vst1q_p8_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1q_p16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.poly16x8x4_t, align 16
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x8x4_t, align 16
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x i16>] [[B]].coerce, [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly16x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly16x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK:   [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK:   [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK:   [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK:   [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8i16.p0i16(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i16* [[TMP15]])
-// CHECK:   ret void
-void test_vst1q_p16_x4(poly16_t *a, poly16x8x4_t b) {
-  vst1q_p16_x4(a, b);
-}
-
 // CHECK-LABEL: @test_vst1q_p64_x4(
 // CHECK:   [[B:%.*]] = alloca %struct.poly64x2x4_t, align 16
 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x4_t, align 16
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[B]], i32 0, i32 0
 // CHECK:   store [4 x <2 x i64>] [[B]].coerce, [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x2x4_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)

 // CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
 // CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
 // CHECK:   call void @llvm.aarch64.neon.st1x4.v2i64.p0i64(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64* [[TMP15]])
 // CHECK:   ret void
 void test_vst1q_p64_x4(poly64_t *a, poly64x2x4_t b) {
   vst1q_p64_x4(a, b);
 }
 
-// CHECK-LABEL: @test_vst1_u8_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint8x8x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x8x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x i8>] [[B]].coerce, [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint8x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint8x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i8* %a)
-// CHECK:   ret void
-void test_vst1_u8_x4(uint8_t *a, uint8x8x4_t b) {
-  vst1_u8_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint16x4x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x4x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x i16>] [[B]].coerce, [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint16x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint16x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4i16.p0i16(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i16* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_u16_x4(uint16_t *a, uint16x4x4_t b) {
-  vst1_u16_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u32_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint32x2x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x2x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <2 x i32>] [[B]].coerce, [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint32x2x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint32x2x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <2 x i32> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x i32>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v2i32.p0i32(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i32* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_u32_x4(uint32_t *a, uint32x2x4_t b) {
-  vst1_u32_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_u64_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.uint64x1x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x1x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <1 x i64>] [[B]].coerce, [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x1x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.uint64x1x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v1i64.p0i64(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_u64_x4(uint64_t *a, uint64x1x4_t b) {
-  vst1_u64_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s8_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int8x8x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int8x8x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x i8>] [[B]].coerce, [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int8x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int8x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i8* %a)
-// CHECK:   ret void
-void test_vst1_s8_x4(int8_t *a, int8x8x4_t b) {
-  vst1_s8_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int16x4x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int16x4x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x i16>] [[B]].coerce, [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int16x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int16x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4i16.p0i16(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i16* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_s16_x4(int16_t *a, int16x4x4_t b) {
-  vst1_s16_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s32_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int32x2x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int32x2x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <2 x i32>] [[B]].coerce, [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int32x2x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int32x2x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i32* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <2 x i32> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x i32>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v2i32.p0i32(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i32* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_s32_x4(int32_t *a, int32x2x4_t b) {
-  vst1_s32_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_s64_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.int64x1x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.int64x1x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <1 x i64>] [[B]].coerce, [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x1x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.int64x1x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v1i64.p0i64(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_s64_x4(int64_t *a, int64x1x4_t b) {
-  vst1_s64_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_f16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.float16x4x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.float16x4x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x half>] [[B]].coerce, [4 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float16x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float16x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast half* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x half> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x half> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x half>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x half>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x half>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x half>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to half*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4f16.p0f16(<4 x half> [[TMP11]], <4 x half> [[TMP12]], <4 x half> [[TMP13]], <4 x half> [[TMP14]], half* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_f16_x4(float16_t *a, float16x4x4_t b) {
-  vst1_f16_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_f32_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.float32x2x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.float32x2x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <2 x float>] [[B]].coerce, [4 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.float32x2x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.float32x2x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast float* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <2 x float> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <2 x float> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x float>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x float>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to float*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v2f32.p0f32(<2 x float> [[TMP11]], <2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]], float* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_f32_x4(float32_t *a, float32x2x4_t b) {
-  vst1_f32_x4(a, b);
-}
-
 // CHECK-LABEL: @test_vst1_f64_x4(
 // CHECK:   [[B:%.*]] = alloca %struct.float64x1x4_t, align 8
 // CHECK:   [[__S1:%.*]] = alloca %struct.float64x1x4_t, align 8
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[B]], i32 0, i32 0
 // CHECK:   store [4 x <1 x double>] [[B]].coerce, [4 x <1 x double>]* [[COERCE_DIVE]], align 8
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x1x4_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.float64x1x4_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)

 // CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x double>
 // CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to double*
 // CHECK:   call void @llvm.aarch64.neon.st1x4.v1f64.p0f64(<1 x double> [[TMP11]], <1 x double> [[TMP12]], <1 x double> [[TMP13]], <1 x double> [[TMP14]], double* [[TMP15]])
 // CHECK:   ret void
 void test_vst1_f64_x4(float64_t *a, float64x1x4_t b) {
   vst1_f64_x4(a, b);
 }
 
-// CHECK-LABEL: @test_vst1_p8_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.poly8x8x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x8x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <8 x i8>] [[B]].coerce, [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly8x8x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly8x8x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i8* %a)
-// CHECK:   ret void
-void test_vst1_p8_x4(poly8_t *a, poly8x8x4_t b) {
-  vst1_p8_x4(a, b);
-}
-
-// CHECK-LABEL: @test_vst1_p16_x4(
-// CHECK:   [[B:%.*]] = alloca %struct.poly16x4x4_t, align 8
-// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x4x4_t, align 8
-// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[B]], i32 0, i32 0
-// CHECK:   store [4 x <4 x i16>] [[B]].coerce, [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly16x4x4_t* [[__S1]] to i8*
-// CHECK:   [[TMP1:%.*]] = bitcast %struct.poly16x4x4_t* [[B]] to i8*
-// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK:   [[TMP2:%.*]] = bitcast i16* %a to i8*
-// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK:   [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK:   [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK:   [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK:   [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK:   [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK:   [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK:   [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK:   [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
-// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
-// CHECK:   [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
-// CHECK:   call void @llvm.aarch64.neon.st1x4.v4i16.p0i16(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i16* [[TMP15]])
-// CHECK:   ret void
-void test_vst1_p16_x4(poly16_t *a, poly16x4x4_t b) {
-  vst1_p16_x4(a, b);
-}
-
 // CHECK-LABEL: @test_vst1_p64_x4(
 // CHECK:   [[B:%.*]] = alloca %struct.poly64x1x4_t, align 8
 // CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x4_t, align 8
 // CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[B]], i32 0, i32 0
 // CHECK:   store [4 x <1 x i64>] [[B]].coerce, [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
 // CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__S1]] to i8*
 // CHECK:   [[TMP1:%.*]] = bitcast %struct.poly64x1x4_t* [[B]] to i8*
 // CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)

cfe/trunk/test/CodeGen/arm-neon-vst.c

+// RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
+// RUN:     -S -disable-O0-optnone -emit-llvm -o - %s | opt -S -mem2reg | \
+// RUN:     FileCheck -check-prefixes=CHECK,CHECK-A64 %s
+// RUN: %clang_cc1 -triple armv8-none-linux-gnueabi -target-feature +neon \
+// RUN:     -S -disable-O0-optnone -emit-llvm -o - %s | opt -S -mem2reg | \
+// RUN:     FileCheck -check-prefixes=CHECK,CHECK-A32 %s
+
+#include <arm_neon.h>
+
+// CHECK-LABEL: @test_vst1_f16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.float16x4x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.float16x4x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x half>] [[B]].coerce, [2 x <4 x half>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x half>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float16x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float16x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast half* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x [[HALF:(half|i16)]]>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to [[HALF]]*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4f16.p0f16(<4 x half> [[TMP7]], <4 x half> [[TMP8]], half* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v4i16(i16* [[TMP9]], <4 x i16> [[TMP7]], <4 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1_f16_x2(float16_t *a, float16x4x2_t b) {
+  vst1_f16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_f16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.float16x4x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.float16x4x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x half>] [[B]].coerce, [3 x <4 x half>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x half>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float16x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float16x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast half* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x half> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to [[HALF]]*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4f16.p0f16(<4 x half> [[TMP9]], <4 x half> [[TMP10]], <4 x half> [[TMP11]], half* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v4i16(i16* [[TMP12]], <4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1_f16_x3(float16_t *a, float16x4x3_t b) {
+  vst1_f16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_f16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.float16x4x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.float16x4x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x half>] [[B]].coerce, [4 x <4 x half>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x half>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float16x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float16x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast half* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x half> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <4 x half> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x [[HALF]]>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to [[HALF]]*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4f16.p0f16(<4 x half> [[TMP11]], <4 x half> [[TMP12]], <4 x half> [[TMP13]], <4 x half> [[TMP14]], half* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v4i16(i16* [[TMP15]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1_f16_x4(float16_t *a, float16x4x4_t b) {
+  vst1_f16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_f32_x2(
+// CHECK: [[B:%.*]] = alloca %struct.float32x2x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.float32x2x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <2 x float>] [[B]].coerce, [2 x <2 x float>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <2 x float>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float32x2x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float32x2x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast float* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to float*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v2f32.p0f32(<2 x float> [[TMP7]], <2 x float> [[TMP8]], float* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0f32.v2f32(float* [[TMP9]], <2 x float> [[TMP7]], <2 x float> [[TMP8]])
+// CHECK: ret void
+void test_vst1_f32_x2(float32_t *a, float32x2x2_t b) {
+  vst1_f32_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_f32_x3(
+// CHECK: [[B:%.*]] = alloca %struct.float32x2x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.float32x2x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <2 x float>] [[B]].coerce, [3 x <2 x float>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <2 x float>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float32x2x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float32x2x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast float* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <2 x float> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x float>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to float*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v2f32.p0f32(<2 x float> [[TMP9]], <2 x float> [[TMP10]], <2 x float> [[TMP11]], float* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0f32.v2f32(float* [[TMP12]], <2 x float> [[TMP9]], <2 x float> [[TMP10]], <2 x float> [[TMP11]])
+// CHECK: ret void
+void test_vst1_f32_x3(float32_t *a, float32x2x3_t b) {
+  vst1_f32_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_f32_x4(
+// CHECK: [[B:%.*]] = alloca %struct.float32x2x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.float32x2x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <2 x float>] [[B]].coerce, [4 x <2 x float>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <2 x float>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float32x2x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float32x2x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast float* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <2 x float> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <2 x float> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x float>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x float>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to float*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v2f32.p0f32(<2 x float> [[TMP11]], <2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]], float* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0f32.v2f32(float* [[TMP15]], <2 x float> [[TMP11]], <2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]])
+// CHECK: ret void
+void test_vst1_f32_x4(float32_t *a, float32x2x4_t b) {
+  vst1_f32_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_p16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.poly16x4x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.poly16x4x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x i16>] [[B]].coerce, [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x i16>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly16x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly16x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4i16.p0i16(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i16* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v4i16(i16* [[TMP9]], <4 x i16> [[TMP7]], <4 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1_p16_x2(poly16_t *a, poly16x4x2_t b) {
+  vst1_p16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_p16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.poly16x4x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.poly16x4x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x i16>] [[B]].coerce, [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x i16>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly16x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly16x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4i16.p0i16(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i16* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v4i16(i16* [[TMP12]], <4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1_p16_x3(poly16_t *a, poly16x4x3_t b) {
+  vst1_p16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_p16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.poly16x4x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.poly16x4x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x i16>] [[B]].coerce, [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x i16>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly16x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly16x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4i16.p0i16(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i16* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v4i16(i16* [[TMP15]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1_p16_x4(poly16_t *a, poly16x4x4_t b) {
+  vst1_p16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_p8_x2(
+// CHECK: [[B:%.*]] = alloca %struct.poly8x8x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.poly8x8x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x i8>] [[B]].coerce, [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x i8>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly8x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly8x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]])
+// CHECK: ret void
+void test_vst1_p8_x2(poly8_t *a, poly8x8x2_t b) {
+  vst1_p8_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_p8_x3(
+// CHECK: [[B:%.*]] = alloca %struct.poly8x8x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.poly8x8x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x i8>] [[B]].coerce, [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x i8>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly8x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly8x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]])
+// CHECK: ret void
+void test_vst1_p8_x3(poly8_t *a, poly8x8x3_t b) {
+  vst1_p8_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_p8_x4(
+// CHECK: [[B:%.*]] = alloca %struct.poly8x8x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.poly8x8x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x i8>] [[B]].coerce, [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x i8>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly8x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly8x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]])
+// CHECK: ret void
+void test_vst1_p8_x4(poly8_t *a, poly8x8x4_t b) {
+  vst1_p8_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int16x4x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int16x4x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x i16>] [[B]].coerce, [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x i16>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int16x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int16x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4i16.p0i16(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i16* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v4i16(i16* [[TMP9]], <4 x i16> [[TMP7]], <4 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1_s16_x2(int16_t *a, int16x4x2_t b) {
+  vst1_s16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int16x4x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int16x4x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x i16>] [[B]].coerce, [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x i16>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int16x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int16x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4i16.p0i16(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i16* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v4i16(i16* [[TMP12]], <4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1_s16_x3(int16_t *a, int16x4x3_t b) {
+  vst1_s16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int16x4x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int16x4x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x i16>] [[B]].coerce, [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x i16>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int16x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int16x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4i16.p0i16(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i16* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v4i16(i16* [[TMP15]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1_s16_x4(int16_t *a, int16x4x4_t b) {
+  vst1_s16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s32_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int32x2x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int32x2x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <2 x i32>] [[B]].coerce, [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <2 x i32>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int32x2x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int32x2x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v2i32.p0i32(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i32* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i32.v2i32(i32* [[TMP9]], <2 x i32> [[TMP7]], <2 x i32> [[TMP8]])
+// CHECK: ret void
+void test_vst1_s32_x2(int32_t *a, int32x2x2_t b) {
+  vst1_s32_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s32_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int32x2x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int32x2x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <2 x i32>] [[B]].coerce, [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <2 x i32>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int32x2x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int32x2x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v2i32.p0i32(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i32* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i32.v2i32(i32* [[TMP12]], <2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]])
+// CHECK: ret void
+void test_vst1_s32_x3(int32_t *a, int32x2x3_t b) {
+  vst1_s32_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s32_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int32x2x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int32x2x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <2 x i32>] [[B]].coerce, [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <2 x i32>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int32x2x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int32x2x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <2 x i32> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x i32>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v2i32.p0i32(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i32* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i32.v2i32(i32* [[TMP15]], <2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]])
+// CHECK: ret void
+void test_vst1_s32_x4(int32_t *a, int32x2x4_t b) {
+  vst1_s32_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s64_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int64x1x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int64x1x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x1x2_t, %struct.int64x1x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <1 x i64>] [[B]].coerce, [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <1 x i64>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int64x1x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int64x1x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int64x1x2_t, %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int64x1x2_t, %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v1i64.p0i64(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i64.v1i64(i64* [[TMP9]], <1 x i64> [[TMP7]], <1 x i64> [[TMP8]])
+// CHECK: ret void
+void test_vst1_s64_x2(int64_t *a, int64x1x2_t b) {
+  vst1_s64_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s64_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int64x1x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int64x1x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <1 x i64>] [[B]].coerce, [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <1 x i64>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int64x1x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int64x1x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int64x1x3_t, %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v1i64.p0i64(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i64.v1i64(i64* [[TMP12]], <1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]])
+// CHECK: ret void
+void test_vst1_s64_x3(int64_t *a, int64x1x3_t b) {
+  vst1_s64_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s64_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int64x1x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int64x1x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <1 x i64>] [[B]].coerce, [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <1 x i64>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int64x1x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int64x1x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int64x1x4_t, %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v1i64.p0i64(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i64.v1i64(i64* [[TMP15]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]])
+// CHECK: ret void
+void test_vst1_s64_x4(int64_t *a, int64x1x4_t b) {
+  vst1_s64_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s8_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int8x8x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int8x8x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x i8>] [[B]].coerce, [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x i8>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int8x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int8x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]])
+// CHECK: ret void
+void test_vst1_s8_x2(int8_t *a, int8x8x2_t b) {
+  vst1_s8_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s8_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int8x8x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int8x8x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x i8>] [[B]].coerce, [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x i8>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int8x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int8x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]])
+// CHECK: ret void
+void test_vst1_s8_x3(int8_t *a, int8x8x3_t b) {
+  vst1_s8_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_s8_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int8x8x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.int8x8x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x i8>] [[B]].coerce, [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x i8>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int8x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int8x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]])
+// CHECK: ret void
+void test_vst1_s8_x4(int8_t *a, int8x8x4_t b) {
+  vst1_s8_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint16x4x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint16x4x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x i16>] [[B]].coerce, [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x i16>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint16x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint16x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4i16.p0i16(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i16* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v4i16(i16* [[TMP9]], <4 x i16> [[TMP7]], <4 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1_u16_x2(uint16_t *a, uint16x4x2_t b) {
+  vst1_u16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint16x4x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint16x4x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x i16>] [[B]].coerce, [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x i16>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint16x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint16x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4i16.p0i16(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i16* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v4i16(i16* [[TMP12]], <4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1_u16_x3(uint16_t *a, uint16x4x3_t b) {
+  vst1_u16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint16x4x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint16x4x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x i16>] [[B]].coerce, [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x i16>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint16x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint16x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4i16.p0i16(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i16* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v4i16(i16* [[TMP15]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1_u16_x4(uint16_t *a, uint16x4x4_t b) {
+  vst1_u16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u32_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint32x2x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint32x2x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <2 x i32>] [[B]].coerce, [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <2 x i32>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint32x2x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint32x2x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v2i32.p0i32(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i32* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i32.v2i32(i32* [[TMP9]], <2 x i32> [[TMP7]], <2 x i32> [[TMP8]])
+// CHECK: ret void
+void test_vst1_u32_x2(uint32_t *a, uint32x2x2_t b) {
+  vst1_u32_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u32_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint32x2x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint32x2x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <2 x i32>] [[B]].coerce, [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <2 x i32>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint32x2x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint32x2x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v2i32.p0i32(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i32* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i32.v2i32(i32* [[TMP12]], <2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]])
+// CHECK: ret void
+void test_vst1_u32_x3(uint32_t *a, uint32x2x3_t b) {
+  vst1_u32_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u32_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint32x2x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint32x2x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <2 x i32>] [[B]].coerce, [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <2 x i32>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint32x2x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint32x2x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <2 x i32> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x i32>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v2i32.p0i32(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i32* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i32.v2i32(i32* [[TMP15]], <2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]])
+// CHECK: ret void
+void test_vst1_u32_x4(uint32_t *a, uint32x2x4_t b) {
+  vst1_u32_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u64_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint64x1x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint64x1x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <1 x i64>] [[B]].coerce, [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <1 x i64>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint64x1x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint64x1x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v1i64.p0i64(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i64.v1i64(i64* [[TMP9]], <1 x i64> [[TMP7]], <1 x i64> [[TMP8]])
+// CHECK: ret void
+void test_vst1_u64_x2(uint64_t *a, uint64x1x2_t b) {
+  vst1_u64_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u64_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint64x1x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint64x1x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <1 x i64>] [[B]].coerce, [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <1 x i64>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint64x1x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint64x1x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v1i64.p0i64(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i64.v1i64(i64* [[TMP12]], <1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]])
+// CHECK: ret void
+void test_vst1_u64_x3(uint64_t *a, uint64x1x3_t b) {
+  vst1_u64_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u64_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint64x1x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint64x1x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <1 x i64>] [[B]].coerce, [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <1 x i64>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint64x1x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint64x1x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK: [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
+// CHECK: [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v1i64.p0i64(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i64.v1i64(i64* [[TMP15]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]])
+// CHECK: ret void
+void test_vst1_u64_x4(uint64_t *a, uint64x1x4_t b) {
+  vst1_u64_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u8_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint8x8x2_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint8x8x2_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x i8>] [[B]].coerce, [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x i8>]* %coerce.dive to [2 x i64]*
+// CHECK-A32: store [2 x i64] %b.coerce, [2 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint8x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint8x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 16, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]])
+// CHECK: ret void
+void test_vst1_u8_x2(uint8_t *a, uint8x8x2_t b) {
+  vst1_u8_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u8_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint8x8x3_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint8x8x3_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x i8>] [[B]].coerce, [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x i8>]* %coerce.dive to [3 x i64]*
+// CHECK-A32: store [3 x i64] %b.coerce, [3 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint8x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint8x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 24, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]])
+// CHECK: ret void
+void test_vst1_u8_x3(uint8_t *a, uint8x8x3_t b) {
+  vst1_u8_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1_u8_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint8x8x4_t, align 8
+// CHECK: [[__S1:%.*]] = alloca %struct.uint8x8x4_t, align 8
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x i8>] [[B]].coerce, [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x i8>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint8x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint8x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i8.v8i8(i8* %a, <8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]])
+// CHECK: ret void
+void test_vst1_u8_x4(uint8_t *a, uint8x8x4_t b) {
+  vst1_u8_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_f16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.float16x8x2_t, align [[QALIGN:(16|8)]]
+// CHECK: [[__S1:%.*]] = alloca %struct.float16x8x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x half>] [[B]].coerce, [2 x <8 x half>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x half>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float16x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float16x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast half* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to [[HALF]]*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8f16.p0f16(<8 x half> [[TMP7]], <8 x half> [[TMP8]], half* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v8i16(i16* [[TMP9]], <8 x i16> [[TMP7]], <8 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_f16_x2(float16_t *a, float16x8x2_t b) {
+  vst1q_f16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_f16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.float16x8x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.float16x8x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x half>] [[B]].coerce, [3 x <8 x half>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x half>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float16x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float16x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast half* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x half> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to [[HALF]]*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8f16.p0f16(<8 x half> [[TMP9]], <8 x half> [[TMP10]], <8 x half> [[TMP11]], half* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v8i16(i16* [[TMP12]], <8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_f16_x3(float16_t *a, float16x8x3_t b) {
+  vst1q_f16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_f16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.float16x8x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.float16x8x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x half>] [[B]].coerce, [4 x <8 x half>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x half>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float16x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float16x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast half* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x half> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <8 x half> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x [[HALF]]>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to [[HALF]]*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8f16.p0f16(<8 x half> [[TMP11]], <8 x half> [[TMP12]], <8 x half> [[TMP13]], <8 x half> [[TMP14]], half* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v8i16(i16* [[TMP15]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_f16_x4(float16_t *a, float16x8x4_t b) {
+  vst1q_f16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_f32_x2(
+// CHECK: [[B:%.*]] = alloca %struct.float32x4x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.float32x4x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x float>] [[B]].coerce, [2 x <4 x float>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x float>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float32x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float32x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast float* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to float*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4f32.p0f32(<4 x float> [[TMP7]], <4 x float> [[TMP8]], float* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0f32.v4f32(float* [[TMP9]], <4 x float> [[TMP7]], <4 x float> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_f32_x2(float32_t *a, float32x4x2_t b) {
+  vst1q_f32_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_f32_x3(
+// CHECK: [[B:%.*]] = alloca %struct.float32x4x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.float32x4x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x float>] [[B]].coerce, [3 x <4 x float>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x float>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float32x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float32x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast float* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <4 x float> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x float>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to float*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4f32.p0f32(<4 x float> [[TMP9]], <4 x float> [[TMP10]], <4 x float> [[TMP11]], float* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0f32.v4f32(float* [[TMP12]], <4 x float> [[TMP9]], <4 x float> [[TMP10]], <4 x float> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_f32_x3(float32_t *a, float32x4x3_t b) {
+  vst1q_f32_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_f32_x4(
+// CHECK: [[B:%.*]] = alloca %struct.float32x4x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.float32x4x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x float>] [[B]].coerce, [4 x <4 x float>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x float>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.float32x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.float32x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast float* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <4 x float> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <4 x float> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x float>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x float>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to float*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4f32.p0f32(<4 x float> [[TMP11]], <4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]], float* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0f32.v4f32(float* [[TMP15]], <4 x float> [[TMP11]], <4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_f32_x4(float32_t *a, float32x4x4_t b) {
+  vst1q_f32_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_p16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.poly16x8x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.poly16x8x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x i16>] [[B]].coerce, [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x i16>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly16x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly16x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8i16.p0i16(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i16* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v8i16(i16* [[TMP9]], <8 x i16> [[TMP7]], <8 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_p16_x2(poly16_t *a, poly16x8x2_t b) {
+  vst1q_p16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_p16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.poly16x8x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.poly16x8x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x i16>] [[B]].coerce, [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x i16>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly16x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly16x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8i16.p0i16(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i16* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v8i16(i16* [[TMP12]], <8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_p16_x3(poly16_t *a, poly16x8x3_t b) {
+  vst1q_p16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_p16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.poly16x8x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.poly16x8x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x i16>] [[B]].coerce, [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x i16>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly16x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly16x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8i16.p0i16(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i16* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v8i16(i16* [[TMP15]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_p16_x4(poly16_t *a, poly16x8x4_t b) {
+  vst1q_p16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_p8_x2(
+// CHECK: [[B:%.*]] = alloca %struct.poly8x16x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.poly8x16x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <16 x i8>] [[B]].coerce, [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <16 x i8>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly8x16x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly8x16x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]])
+// CHECK: ret void
+void test_vst1q_p8_x2(poly8_t *a, poly8x16x2_t b) {
+  vst1q_p8_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_p8_x3(
+// CHECK: [[B:%.*]] = alloca %struct.poly8x16x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.poly8x16x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <16 x i8>] [[B]].coerce, [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <16 x i8>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly8x16x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly8x16x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]])
+// CHECK: ret void
+void test_vst1q_p8_x3(poly8_t *a, poly8x16x3_t b) {
+  vst1q_p8_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_p8_x4(
+// CHECK: [[B:%.*]] = alloca %struct.poly8x16x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.poly8x16x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <16 x i8>] [[B]].coerce, [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <16 x i8>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.poly8x16x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.poly8x16x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]])
+// CHECK: ret void
+void test_vst1q_p8_x4(poly8_t *a, poly8x16x4_t b) {
+  vst1q_p8_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int16x8x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int16x8x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x i16>] [[B]].coerce, [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x i16>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int16x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int16x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8i16.p0i16(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i16* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v8i16(i16* [[TMP9]], <8 x i16> [[TMP7]], <8 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_s16_x2(int16_t *a, int16x8x2_t b) {
+  vst1q_s16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int16x8x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int16x8x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x i16>] [[B]].coerce, [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x i16>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int16x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int16x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8i16.p0i16(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i16* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v8i16(i16* [[TMP12]], <8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_s16_x3(int16_t *a, int16x8x3_t b) {
+  vst1q_s16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int16x8x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int16x8x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x i16>] [[B]].coerce, [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x i16>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int16x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int16x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8i16.p0i16(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i16* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v8i16(i16* [[TMP15]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_s16_x4(int16_t *a, int16x8x4_t b) {
+  vst1q_s16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s32_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int32x4x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int32x4x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x i32>] [[B]].coerce, [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x i32>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int32x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int32x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4i32.p0i32(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i32* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i32.v4i32(i32* [[TMP9]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_s32_x2(int32_t *a, int32x4x2_t b) {
+  vst1q_s32_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s32_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int32x4x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int32x4x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x i32>] [[B]].coerce, [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x i32>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int32x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int32x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4i32.p0i32(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i32* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i32.v4i32(i32* [[TMP12]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_s32_x3(int32_t *a, int32x4x3_t b) {
+  vst1q_s32_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s32_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int32x4x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int32x4x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x i32>] [[B]].coerce, [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x i32>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int32x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int32x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <4 x i32> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x i32>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4i32.p0i32(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i32* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i32.v4i32(i32* [[TMP15]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_s32_x4(int32_t *a, int32x4x4_t b) {
+  vst1q_s32_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s64_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int64x2x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int64x2x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x2x2_t, %struct.int64x2x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <2 x i64>] [[B]].coerce, [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <2 x i64>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int64x2x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int64x2x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int64x2x2_t, %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int64x2x2_t, %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v2i64.p0i64(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i64.v2i64(i64* [[TMP9]], <2 x i64> [[TMP7]], <2 x i64> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_s64_x2(int64_t *a, int64x2x2_t b) {
+  vst1q_s64_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s64_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int64x2x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int64x2x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <2 x i64>] [[B]].coerce, [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <2 x i64>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int64x2x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int64x2x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int64x2x3_t, %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v2i64.p0i64(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i64.v2i64(i64* [[TMP12]], <2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_s64_x3(int64_t *a, int64x2x3_t b) {
+  vst1q_s64_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s64_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int64x2x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int64x2x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <2 x i64>] [[B]].coerce, [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <2 x i64>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int64x2x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int64x2x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int64x2x4_t, %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v2i64.p0i64(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i64.v2i64(i64* [[TMP15]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_s64_x4(int64_t *a, int64x2x4_t b) {
+  vst1q_s64_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s8_x2(
+// CHECK: [[B:%.*]] = alloca %struct.int8x16x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int8x16x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <16 x i8>] [[B]].coerce, [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <16 x i8>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int8x16x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int8x16x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]])
+// CHECK: ret void
+void test_vst1q_s8_x2(int8_t *a, int8x16x2_t b) {
+  vst1q_s8_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s8_x3(
+// CHECK: [[B:%.*]] = alloca %struct.int8x16x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int8x16x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <16 x i8>] [[B]].coerce, [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <16 x i8>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int8x16x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int8x16x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]])
+// CHECK: ret void
+void test_vst1q_s8_x3(int8_t *a, int8x16x3_t b) {
+  vst1q_s8_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_s8_x4(
+// CHECK: [[B:%.*]] = alloca %struct.int8x16x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.int8x16x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <16 x i8>] [[B]].coerce, [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <16 x i8>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.int8x16x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.int8x16x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.int8x16x4_t, %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]])
+// CHECK: ret void
+void test_vst1q_s8_x4(int8_t *a, int8x16x4_t b) {
+  vst1q_s8_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u16_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint16x8x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint16x8x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <8 x i16>] [[B]].coerce, [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <8 x i16>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint16x8x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint16x8x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v8i16.p0i16(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i16* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i16.v8i16(i16* [[TMP9]], <8 x i16> [[TMP7]], <8 x i16> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_u16_x2(uint16_t *a, uint16x8x2_t b) {
+  vst1q_u16_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u16_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint16x8x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint16x8x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <8 x i16>] [[B]].coerce, [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <8 x i16>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint16x8x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint16x8x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v8i16.p0i16(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i16* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i16.v8i16(i16* [[TMP12]], <8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_u16_x3(uint16_t *a, uint16x8x3_t b) {
+  vst1q_u16_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u16_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint16x8x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint16x8x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <8 x i16>] [[B]].coerce, [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <8 x i16>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint16x8x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint16x8x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i16* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i16*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v8i16.p0i16(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i16* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i16.v8i16(i16* [[TMP15]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_u16_x4(uint16_t *a, uint16x8x4_t b) {
+  vst1q_u16_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u32_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint32x4x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint32x4x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <4 x i32>] [[B]].coerce, [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <4 x i32>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint32x4x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint32x4x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v4i32.p0i32(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i32* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i32.v4i32(i32* [[TMP9]], <4 x i32> [[TMP7]], <4 x i32> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_u32_x2(uint32_t *a, uint32x4x2_t b) {
+  vst1q_u32_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u32_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint32x4x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint32x4x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <4 x i32>] [[B]].coerce, [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <4 x i32>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint32x4x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint32x4x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v4i32.p0i32(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i32* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i32.v4i32(i32* [[TMP12]], <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_u32_x3(uint32_t *a, uint32x4x3_t b) {
+  vst1q_u32_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u32_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint32x4x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint32x4x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <4 x i32>] [[B]].coerce, [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <4 x i32>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint32x4x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint32x4x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i32* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <4 x i32> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x i32>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i32*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v4i32.p0i32(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i32* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i32.v4i32(i32* [[TMP15]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_u32_x4(uint32_t *a, uint32x4x4_t b) {
+  vst1q_u32_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u64_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint64x2x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint64x2x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <2 x i64>] [[B]].coerce, [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <2 x i64>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint64x2x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint64x2x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
+// CHECK-DAG: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
+// CHECK-DAG: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v2i64.p0i64(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64* [[TMP9]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i64.v2i64(i64* [[TMP9]], <2 x i64> [[TMP7]], <2 x i64> [[TMP8]])
+// CHECK: ret void
+void test_vst1q_u64_x2(uint64_t *a, uint64x2x2_t b) {
+  vst1q_u64_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u64_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint64x2x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint64x2x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <2 x i64>] [[B]].coerce, [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <2 x i64>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint64x2x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint64x2x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
+// CHECK-DAG: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
+// CHECK-DAG: [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v2i64.p0i64(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64* [[TMP12]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i64.v2i64(i64* [[TMP12]], <2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]])
+// CHECK: ret void
+void test_vst1q_u64_x3(uint64_t *a, uint64x2x3_t b) {
+  vst1q_u64_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u64_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint64x2x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint64x2x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <2 x i64>] [[B]].coerce, [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <2 x i64>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint64x2x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint64x2x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK: [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
+// CHECK-DAG: [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
+// CHECK-DAG: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
+// CHECK-DAG: [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
+// CHECK-DAG: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
+// CHECK-DAG: [[TMP15:%.*]] = bitcast i8* [[TMP2]] to i64*
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v2i64.p0i64(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64* [[TMP15]])
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i64.v2i64(i64* [[TMP15]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]])
+// CHECK: ret void
+void test_vst1q_u64_x4(uint64_t *a, uint64x2x4_t b) {
+  vst1q_u64_x4(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u8_x2(
+// CHECK: [[B:%.*]] = alloca %struct.uint8x16x2_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint8x16x2_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [2 x <16 x i8>] [[B]].coerce, [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [2 x <16 x i8>]* %coerce.dive to [4 x i64]*
+// CHECK-A32: store [4 x i64] %b.coerce, [4 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint8x16x2_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint8x16x2_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 32, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x2.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x2.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]])
+// CHECK: ret void
+void test_vst1q_u8_x2(uint8_t *a, uint8x16x2_t b) {
+  vst1q_u8_x2(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u8_x3(
+// CHECK: [[B:%.*]] = alloca %struct.uint8x16x3_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint8x16x3_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [3 x <16 x i8>] [[B]].coerce, [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [3 x <16 x i8>]* %coerce.dive to [6 x i64]*
+// CHECK-A32: store [6 x i64] %b.coerce, [6 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint8x16x3_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint8x16x3_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 48, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x3.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x3.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]])
+// CHECK: ret void
+void test_vst1q_u8_x3(uint8_t *a, uint8x16x3_t b) {
+  vst1q_u8_x3(a, b);
+}
+
+// CHECK-LABEL: @test_vst1q_u8_x4(
+// CHECK: [[B:%.*]] = alloca %struct.uint8x16x4_t, align [[QALIGN]]
+// CHECK: [[__S1:%.*]] = alloca %struct.uint8x16x4_t, align [[QALIGN]]
+// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[B]], i32 0, i32 0
+// CHECK-A64: store [4 x <16 x i8>] [[B]].coerce, [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
+// CHECK-A32: [[COERCE_DIVE_TMP:%.*]] = bitcast [4 x <16 x i8>]* %coerce.dive to [8 x i64]*
+// CHECK-A32: store [8 x i64] %b.coerce, [8 x i64]* [[COERCE_DIVE_TMP]], align 8
+// CHECK: [[TMP0:%.*]] = bitcast %struct.uint8x16x4_t* [[__S1]] to i8*
+// CHECK: [[TMP1:%.*]] = bitcast %struct.uint8x16x4_t* [[B]] to i8*
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.{{i64|i32}}(i8* align [[QALIGN]] [[TMP0]], i8* align [[QALIGN]] [[TMP1]], {{i64|i32}} 64, i1 false)
+// CHECK: [[VAL:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], {{i64|i32}} 0, {{i64|i32}} 0
+// CHECK: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align [[QALIGN]]
+// CHECK: [[VAL1:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], {{i64|i32}} 0, {{i64|i32}} 1
+// CHECK: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align [[QALIGN]]
+// CHECK: [[VAL3:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], {{i64|i32}} 0, {{i64|i32}} 2
+// CHECK: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align [[QALIGN]]
+// CHECK: [[VAL5:%.*]] = getelementptr inbounds %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
+// CHECK: [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], {{i64|i32}} 0, {{i64|i32}} 3
+// CHECK: [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align [[QALIGN]]
+// CHECK-A64: call void @llvm.aarch64.neon.st1x4.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i8* %a)
+// CHECK-A32: call void @llvm.arm.neon.vst1x4.p0i8.v16i8(i8* %a, <16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]])
+// CHECK: ret void
+void test_vst1q_u8_x4(uint8_t *a, uint8x16x4_t b) {
+  vst1q_u8_x4(a, b);
+}