commit 2cb8c10342ee5d040725abb1166feb92a64c7df6
Author: Dávid Bolvanský <david.bolvansky@gmail.com>
Date:   Thu Apr 8 17:28:57 2021 +0200

    Revert "Reduce the number of attributes attached to each function"
    
    This reverts commit 053dc95839b3b8a36db46f8c419e36e632e989cd. It causes perf regressions - see discussion in D97116.

diff --git a/clang/lib/CodeGen/CodeGenFunction.cpp b/clang/lib/CodeGen/CodeGenFunction.cpp
index 600312e15ef4..5ff969dbf285 100644
--- a/clang/lib/CodeGen/CodeGenFunction.cpp
+++ b/clang/lib/CodeGen/CodeGenFunction.cpp
@@ -493,8 +493,7 @@ void CodeGenFunction::FinishFunction(SourceLocation EndLoc) {
   // 4. Width of vector arguments and return types for this function.
   // 5. Width of vector aguments and return types for functions called by this
   //    function.
-  if (LargestVectorWidth)
-    CurFn->addFnAttr("min-legal-vector-width", llvm::utostr(LargestVectorWidth));
+  CurFn->addFnAttr("min-legal-vector-width", llvm::utostr(LargestVectorWidth));
 
   // Add vscale attribute if appropriate.
   if (getLangOpts().ArmSveVectorBits) {
diff --git a/clang/test/CodeGen/aarch64-neon-ldst-one.c b/clang/test/CodeGen/aarch64-neon-ldst-one.c
index e4fce996a1f3..db352fecf221 100644
--- a/clang/test/CodeGen/aarch64-neon-ldst-one.c
+++ b/clang/test/CodeGen/aarch64-neon-ldst-one.c
@@ -1,7341 +1,6803 @@
-// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
 // RUN:  -disable-O0-optnone -fallow-half-arguments-and-returns -emit-llvm -o - %s \
 // RUN: | opt -S -mem2reg | FileCheck %s
 
 #include <arm_neon.h>
 
-// CHECK-LABEL: @test_vld1q_dup_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[TMP0]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP1]], <16 x i32> zeroinitializer
-// CHECK-NEXT:    ret <16 x i8> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vld1q_dup_u8(i8* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[TMP0]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP1]], <16 x i32> zeroinitializer
+// CHECK:   ret <16 x i8> [[LANE]]
 uint8x16_t test_vld1q_dup_u8(uint8_t  *a) {
   return vld1q_dup_u8(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <8 x i16> undef, i16 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x i16> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vld1q_dup_u16(i16* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP2:%.*]] = load i16, i16* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <8 x i16> undef, i16 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x i16> [[LANE]]
 uint16x8_t test_vld1q_dup_u16(uint16_t  *a) {
   return vld1q_dup_u16(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32> undef, i32 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x i32> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i32> @test_vld1q_dup_u32(i32* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP2:%.*]] = load i32, i32* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i32> undef, i32 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x i32> [[LANE]]
 uint32x4_t test_vld1q_dup_u32(uint32_t  *a) {
   return vld1q_dup_u32(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i64> undef, i64 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x i64> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vld1q_dup_u64(i64* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP2:%.*]] = load i64, i64* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x i64> undef, i64 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x i64> [[LANE]]
 uint64x2_t test_vld1q_dup_u64(uint64_t  *a) {
   return vld1q_dup_u64(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[TMP0]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP1]], <16 x i32> zeroinitializer
-// CHECK-NEXT:    ret <16 x i8> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vld1q_dup_s8(i8* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[TMP0]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP1]], <16 x i32> zeroinitializer
+// CHECK:   ret <16 x i8> [[LANE]]
 int8x16_t test_vld1q_dup_s8(int8_t  *a) {
   return vld1q_dup_s8(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <8 x i16> undef, i16 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x i16> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vld1q_dup_s16(i16* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP2:%.*]] = load i16, i16* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <8 x i16> undef, i16 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x i16> [[LANE]]
 int16x8_t test_vld1q_dup_s16(int16_t  *a) {
   return vld1q_dup_s16(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32> undef, i32 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x i32> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i32> @test_vld1q_dup_s32(i32* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP2:%.*]] = load i32, i32* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i32> undef, i32 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x i32> [[LANE]]
 int32x4_t test_vld1q_dup_s32(int32_t  *a) {
   return vld1q_dup_s32(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i64> undef, i64 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x i64> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vld1q_dup_s64(i64* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP2:%.*]] = load i64, i64* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x i64> undef, i64 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x i64> [[LANE]]
 int64x2_t test_vld1q_dup_s64(int64_t  *a) {
   return vld1q_dup_s64(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to half*
-// CHECK-NEXT:    [[TMP2:%.*]] = load half, half* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <8 x half> undef, half [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x half> [[TMP3]], <8 x half> [[TMP3]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x half> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x half> @test_vld1q_dup_f16(half* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast half* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to half*
+// CHECK:   [[TMP2:%.*]] = load half, half* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <8 x half> undef, half [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x half> [[TMP3]], <8 x half> [[TMP3]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x half> [[LANE]]
 float16x8_t test_vld1q_dup_f16(float16_t  *a) {
   return vld1q_dup_f16(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to float*
-// CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[TMP1]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x float> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x float> @test_vld1q_dup_f32(float* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast float* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to float*
+// CHECK:   [[TMP2:%.*]] = load float, float* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x float> undef, float [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x float> [[LANE]]
 float32x4_t test_vld1q_dup_f32(float32_t  *a) {
   return vld1q_dup_f32(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to double*
-// CHECK-NEXT:    [[TMP2:%.*]] = load double, double* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x double> [[TMP3]], <2 x double> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x double> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x double> @test_vld1q_dup_f64(double* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to double*
+// CHECK:   [[TMP2:%.*]] = load double, double* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x double> [[TMP3]], <2 x double> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x double> [[LANE]]
 float64x2_t test_vld1q_dup_f64(float64_t  *a) {
   return vld1q_dup_f64(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[TMP0]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP1]], <16 x i32> zeroinitializer
-// CHECK-NEXT:    ret <16 x i8> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vld1q_dup_p8(i8* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[TMP0]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP1]], <16 x i32> zeroinitializer
+// CHECK:   ret <16 x i8> [[LANE]]
 poly8x16_t test_vld1q_dup_p8(poly8_t  *a) {
   return vld1q_dup_p8(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <8 x i16> undef, i16 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x i16> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vld1q_dup_p16(i16* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP2:%.*]] = load i16, i16* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <8 x i16> undef, i16 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x i16> [[TMP3]], <8 x i16> [[TMP3]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x i16> [[LANE]]
 poly16x8_t test_vld1q_dup_p16(poly16_t  *a) {
   return vld1q_dup_p16(a);
 }
 
-// CHECK-LABEL: @test_vld1q_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i64> undef, i64 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x i64> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vld1q_dup_p64(i64* %a) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP2:%.*]] = load i64, i64* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x i64> undef, i64 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x i64> [[LANE]]
 poly64x2_t test_vld1q_dup_p64(poly64_t  *a) {
   return vld1q_dup_p64(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i8> undef, i8 [[TMP0]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP1]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x i8> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i8> @test_vld1_dup_u8(i8* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[TMP1:%.*]] = insertelement <8 x i8> undef, i8 [[TMP0]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP1]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x i8> [[LANE]]
 uint8x8_t test_vld1_dup_u8(uint8_t  *a) {
   return vld1_dup_u8(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP3]], <4 x i16> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x i16> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i16> @test_vld1_dup_u16(i16* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP2:%.*]] = load i16, i16* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x i16> [[TMP3]], <4 x i16> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x i16> [[LANE]]
 uint16x4_t test_vld1_dup_u16(uint16_t  *a) {
   return vld1_dup_u16(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> undef, i32 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x i32> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i32> @test_vld1_dup_u32(i32* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP2:%.*]] = load i32, i32* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x i32> undef, i32 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x i32> [[LANE]]
 uint32x2_t test_vld1_dup_u32(uint32_t  *a) {
   return vld1_dup_u32(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x i64> [[TMP3]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
-// CHECK-NEXT:    ret <1 x i64> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x i64> @test_vld1_dup_u64(i64* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP2:%.*]] = load i64, i64* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <1 x i64> [[TMP3]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
+// CHECK:   ret <1 x i64> [[LANE]]
 uint64x1_t test_vld1_dup_u64(uint64_t  *a) {
   return vld1_dup_u64(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i8> undef, i8 [[TMP0]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP1]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x i8> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i8> @test_vld1_dup_s8(i8* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[TMP1:%.*]] = insertelement <8 x i8> undef, i8 [[TMP0]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP1]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x i8> [[LANE]]
 int8x8_t test_vld1_dup_s8(int8_t  *a) {
   return vld1_dup_s8(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP3]], <4 x i16> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x i16> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i16> @test_vld1_dup_s16(i16* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP2:%.*]] = load i16, i16* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x i16> [[TMP3]], <4 x i16> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x i16> [[LANE]]
 int16x4_t test_vld1_dup_s16(int16_t  *a) {
   return vld1_dup_s16(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> undef, i32 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x i32> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i32> @test_vld1_dup_s32(i32* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP2:%.*]] = load i32, i32* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x i32> undef, i32 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x i32> [[LANE]]
 int32x2_t test_vld1_dup_s32(int32_t  *a) {
   return vld1_dup_s32(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x i64> [[TMP3]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
-// CHECK-NEXT:    ret <1 x i64> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x i64> @test_vld1_dup_s64(i64* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP2:%.*]] = load i64, i64* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <1 x i64> [[TMP3]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
+// CHECK:   ret <1 x i64> [[LANE]]
 int64x1_t test_vld1_dup_s64(int64_t  *a) {
   return vld1_dup_s64(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to half*
-// CHECK-NEXT:    [[TMP2:%.*]] = load half, half* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x half> undef, half [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x half> [[TMP3]], <4 x half> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x half> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x half> @test_vld1_dup_f16(half* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast half* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to half*
+// CHECK:   [[TMP2:%.*]] = load half, half* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x half> undef, half [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x half> [[TMP3]], <4 x half> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x half> [[LANE]]
 float16x4_t test_vld1_dup_f16(float16_t  *a) {
   return vld1_dup_f16(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to float*
-// CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[TMP1]], align 4
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x float> undef, float [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP3]], <2 x i32> zeroinitializer
-// CHECK-NEXT:    ret <2 x float> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x float> @test_vld1_dup_f32(float* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast float* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to float*
+// CHECK:   [[TMP2:%.*]] = load float, float* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <2 x float> undef, float [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP3]], <2 x i32> zeroinitializer
+// CHECK:   ret <2 x float> [[LANE]]
 float32x2_t test_vld1_dup_f32(float32_t  *a) {
   return vld1_dup_f32(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to double*
-// CHECK-NEXT:    [[TMP2:%.*]] = load double, double* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <1 x double> undef, double [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
-// CHECK-NEXT:    ret <1 x double> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vld1_dup_f64(double* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to double*
+// CHECK:   [[TMP2:%.*]] = load double, double* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <1 x double> undef, double [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
+// CHECK:   ret <1 x double> [[LANE]]
 float64x1_t test_vld1_dup_f64(float64_t  *a) {
   return vld1_dup_f64(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i8> undef, i8 [[TMP0]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP1]], <8 x i32> zeroinitializer
-// CHECK-NEXT:    ret <8 x i8> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i8> @test_vld1_dup_p8(i8* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[TMP1:%.*]] = insertelement <8 x i8> undef, i8 [[TMP0]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP1]], <8 x i32> zeroinitializer
+// CHECK:   ret <8 x i8> [[LANE]]
 poly8x8_t test_vld1_dup_p8(poly8_t  *a) {
   return vld1_dup_p8(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <4 x i16> [[TMP3]], <4 x i16> [[TMP3]], <4 x i32> zeroinitializer
-// CHECK-NEXT:    ret <4 x i16> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i16> @test_vld1_dup_p16(i16* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP2:%.*]] = load i16, i16* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <4 x i16> [[TMP3]], <4 x i16> [[TMP3]], <4 x i32> zeroinitializer
+// CHECK:   ret <4 x i16> [[LANE]]
 poly16x4_t test_vld1_dup_p16(poly16_t  *a) {
   return vld1_dup_p16(a);
 }
 
-// CHECK-LABEL: @test_vld1_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i64, i64* [[TMP1]], align 8
-// CHECK-NEXT:    [[TMP3:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i32 0
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x i64> [[TMP3]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
-// CHECK-NEXT:    ret <1 x i64> [[LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x i64> @test_vld1_dup_p64(i64* %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP2:%.*]] = load i64, i64* [[TMP1]]
+// CHECK:   [[TMP3:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i32 0
+// CHECK:   [[LANE:%.*]] = shufflevector <1 x i64> [[TMP3]], <1 x i64> [[TMP3]], <1 x i32> zeroinitializer
+// CHECK:   ret <1 x i64> [[LANE]]
 poly64x1_t test_vld1_dup_p64(poly64_t  *a) {
   return vld1_dup_p64(a);
 }
 
-// CHECK-LABEL: @test_vld2q_dup_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X2X2_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.uint64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT64X2X2_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x2x2_t @test_vld2q_dup_u64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.uint64x2x2_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.uint64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint64x2x2_t [[TMP6]]
 uint64x2x2_t test_vld2q_dup_u64(uint64_t  *a) {
   return vld2q_dup_u64(a);
 }
 
-// CHECK-LABEL: @test_vld2q_dup_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X2X2_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.int64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT64X2X2_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x2x2_t @test_vld2q_dup_s64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.int64x2x2_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.int64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.int64x2x2_t, %struct.int64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int64x2x2_t [[TMP6]]
 int64x2x2_t test_vld2q_dup_s64(int64_t  *a) {
   return vld2q_dup_s64(a);
 }
 
-// CHECK-LABEL: @test_vld2q_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
-// CHECK-NEXT:    [[VLD2:%.*]] = call { <2 x double>, <2 x double> } @llvm.aarch64.neon.ld2r.v2f64.p0f64(double* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x double>, <2 x double> }*
-// CHECK-NEXT:    store { <2 x double>, <2 x double> } [[VLD2]], { <2 x double>, <2 x double> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.float64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X2X2_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x2x2_t @test_vld2q_dup_f64(double* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.float64x2x2_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
+// CHECK:   [[VLD2:%.*]] = call { <2 x double>, <2 x double> } @llvm.aarch64.neon.ld2r.v2f64.p0f64(double* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x double>, <2 x double> }*
+// CHECK:   store { <2 x double>, <2 x double> } [[VLD2]], { <2 x double>, <2 x double> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.float64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.float64x2x2_t, %struct.float64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float64x2x2_t [[TMP6]]
 float64x2x2_t test_vld2q_dup_f64(float64_t  *a) {
   return vld2q_dup_f64(a);
 }
 
-// CHECK-LABEL: @test_vld2q_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X2X2_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.poly64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY64X2X2_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x2x2_t @test_vld2q_dup_p64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly64x2x2_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD2:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2]], { <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 32, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly64x2x2_t [[TMP6]]
 poly64x2x2_t test_vld2q_dup_p64(poly64_t  *a) {
   return vld2q_dup_p64(a);
 }
 
-// CHECK-LABEL: @test_vld2_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
-// CHECK-NEXT:    [[VLD2:%.*]] = call { <1 x double>, <1 x double> } @llvm.aarch64.neon.ld2r.v1f64.p0f64(double* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x double>, <1 x double> }*
-// CHECK-NEXT:    store { <1 x double>, <1 x double> } [[VLD2]], { <1 x double>, <1 x double> }* [[TMP3]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.float64x1x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X1X2_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x1x2_t @test_vld2_dup_f64(double* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x1x2_t, align 8
+// CHECK:   [[__RET:%.*]] = alloca %struct.float64x1x2_t, align 8
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
+// CHECK:   [[VLD2:%.*]] = call { <1 x double>, <1 x double> } @llvm.aarch64.neon.ld2r.v1f64.p0f64(double* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x double>, <1 x double> }*
+// CHECK:   store { <1 x double>, <1 x double> } [[VLD2]], { <1 x double>, <1 x double> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.float64x1x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 16, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.float64x1x2_t, %struct.float64x1x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float64x1x2_t [[TMP6]]
 float64x1x2_t test_vld2_dup_f64(float64_t  *a) {
   return vld2_dup_f64(a);
 }
 
-// CHECK-LABEL: @test_vld2_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X1X2_T]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD2:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2r.v1i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64> } [[VLD2]], { <1 x i64>, <1 x i64> }* [[TMP3]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.poly64x1x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY64X1X2_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x1x2_t @test_vld2_dup_p64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x2_t, align 8
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly64x1x2_t, align 8
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD2:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2r.v1i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64> } [[VLD2]], { <1 x i64>, <1 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x1x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 16, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly64x1x2_t [[TMP6]]
 poly64x1x2_t test_vld2_dup_p64(poly64_t  *a) {
   return vld2_dup_p64(a);
 }
 
-// CHECK-LABEL: @test_vld3q_dup_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X2X3_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.uint64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT64X2X3_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x2x3_t @test_vld3q_dup_u64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.uint64x2x3_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.uint64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint64x2x3_t [[TMP6]]
 uint64x2x3_t test_vld3q_dup_u64(uint64_t  *a) {
   return vld3q_dup_u64(a);
   // [{{x[0-9]+|sp}}]
 }
 
-// CHECK-LABEL: @test_vld3q_dup_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X2X3_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.int64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT64X2X3_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x2x3_t @test_vld3q_dup_s64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.int64x2x3_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.int64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.int64x2x3_t, %struct.int64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int64x2x3_t [[TMP6]]
 int64x2x3_t test_vld3q_dup_s64(int64_t  *a) {
   return vld3q_dup_s64(a);
   // [{{x[0-9]+|sp}}]
 }
 
-// CHECK-LABEL: @test_vld3q_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
-// CHECK-NEXT:    [[VLD3:%.*]] = call { <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld3r.v2f64.p0f64(double* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x double>, <2 x double>, <2 x double> }*
-// CHECK-NEXT:    store { <2 x double>, <2 x double>, <2 x double> } [[VLD3]], { <2 x double>, <2 x double>, <2 x double> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.float64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X2X3_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x2x3_t @test_vld3q_dup_f64(double* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.float64x2x3_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
+// CHECK:   [[VLD3:%.*]] = call { <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld3r.v2f64.p0f64(double* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x double>, <2 x double>, <2 x double> }*
+// CHECK:   store { <2 x double>, <2 x double>, <2 x double> } [[VLD3]], { <2 x double>, <2 x double>, <2 x double> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.float64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.float64x2x3_t, %struct.float64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float64x2x3_t [[TMP6]]
 float64x2x3_t test_vld3q_dup_f64(float64_t  *a) {
   return vld3q_dup_f64(a);
   // [{{x[0-9]+|sp}}]
 }
 
-// CHECK-LABEL: @test_vld3q_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X2X3_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.poly64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY64X2X3_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x2x3_t @test_vld3q_dup_p64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly64x2x3_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD3:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 48, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly64x2x3_t [[TMP6]]
 poly64x2x3_t test_vld3q_dup_p64(poly64_t  *a) {
   return vld3q_dup_p64(a);
   // [{{x[0-9]+|sp}}]
 }
 
-// CHECK-LABEL: @test_vld3_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
-// CHECK-NEXT:    [[VLD3:%.*]] = call { <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld3r.v1f64.p0f64(double* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x double>, <1 x double>, <1 x double> }*
-// CHECK-NEXT:    store { <1 x double>, <1 x double>, <1 x double> } [[VLD3]], { <1 x double>, <1 x double>, <1 x double> }* [[TMP3]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.float64x1x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X1X3_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x1x3_t @test_vld3_dup_f64(double* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x1x3_t, align 8
+// CHECK:   [[__RET:%.*]] = alloca %struct.float64x1x3_t, align 8
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
+// CHECK:   [[VLD3:%.*]] = call { <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld3r.v1f64.p0f64(double* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x double>, <1 x double>, <1 x double> }*
+// CHECK:   store { <1 x double>, <1 x double>, <1 x double> } [[VLD3]], { <1 x double>, <1 x double>, <1 x double> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.float64x1x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 24, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.float64x1x3_t, %struct.float64x1x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float64x1x3_t [[TMP6]]
 float64x1x3_t test_vld3_dup_f64(float64_t  *a) {
   return vld3_dup_f64(a);
   // [{{x[0-9]+|sp}}]
 }
 
-// CHECK-LABEL: @test_vld3_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X1X3_T]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD3:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3r.v1i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP3]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.poly64x1x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY64X1X3_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x1x3_t @test_vld3_dup_p64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x3_t, align 8
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly64x1x3_t, align 8
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD3:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3r.v1i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x1x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 24, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly64x1x3_t [[TMP6]]
 poly64x1x3_t test_vld3_dup_p64(poly64_t  *a) {
   return vld3_dup_p64(a);
   // [{{x[0-9]+|sp}}]
 }
 
-// CHECK-LABEL: @test_vld4q_dup_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X2X4_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.uint64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT64X2X4_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x2x4_t @test_vld4q_dup_u64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.uint64x2x4_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.uint64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint64x2x4_t [[TMP6]]
 uint64x2x4_t test_vld4q_dup_u64(uint64_t  *a) {
   return vld4q_dup_u64(a);
 }
 
-// CHECK-LABEL: @test_vld4q_dup_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X2X4_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.int64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT64X2X4_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x2x4_t @test_vld4q_dup_s64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.int64x2x4_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.int64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.int64x2x4_t, %struct.int64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int64x2x4_t [[TMP6]]
 int64x2x4_t test_vld4q_dup_s64(int64_t  *a) {
   return vld4q_dup_s64(a);
 }
 
-// CHECK-LABEL: @test_vld4q_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
-// CHECK-NEXT:    [[VLD4:%.*]] = call { <2 x double>, <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld4r.v2f64.p0f64(double* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x double>, <2 x double>, <2 x double>, <2 x double> }*
-// CHECK-NEXT:    store { <2 x double>, <2 x double>, <2 x double>, <2 x double> } [[VLD4]], { <2 x double>, <2 x double>, <2 x double>, <2 x double> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.float64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X2X4_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x2x4_t @test_vld4q_dup_f64(double* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.float64x2x4_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
+// CHECK:   [[VLD4:%.*]] = call { <2 x double>, <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld4r.v2f64.p0f64(double* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x double>, <2 x double>, <2 x double>, <2 x double> }*
+// CHECK:   store { <2 x double>, <2 x double>, <2 x double>, <2 x double> } [[VLD4]], { <2 x double>, <2 x double>, <2 x double>, <2 x double> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.float64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.float64x2x4_t, %struct.float64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float64x2x4_t [[TMP6]]
 float64x2x4_t test_vld4q_dup_f64(float64_t  *a) {
   return vld4q_dup_f64(a);
 }
 
-// CHECK-LABEL: @test_vld4q_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X2X4_T]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4r.v2i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.poly64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY64X2X4_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x2x4_t @test_vld4q_dup_p64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly64x2x4_t, align 16
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD4:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4r.v2i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP4]], i8* align 16 [[TMP5]], i64 64, i1 false)
+// CHECK:   [[TMP6:%.*]] = load %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly64x2x4_t [[TMP6]]
 poly64x2x4_t test_vld4q_dup_p64(poly64_t  *a) {
   return vld4q_dup_p64(a);
 }
 
-// CHECK-LABEL: @test_vld4_dup_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
-// CHECK-NEXT:    [[VLD4:%.*]] = call { <1 x double>, <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld4r.v1f64.p0f64(double* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x double>, <1 x double>, <1 x double>, <1 x double> }*
-// CHECK-NEXT:    store { <1 x double>, <1 x double>, <1 x double>, <1 x double> } [[VLD4]], { <1 x double>, <1 x double>, <1 x double>, <1 x double> }* [[TMP3]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.float64x1x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X1X4_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x1x4_t @test_vld4_dup_f64(double* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x1x4_t, align 8
+// CHECK:   [[__RET:%.*]] = alloca %struct.float64x1x4_t, align 8
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to double*
+// CHECK:   [[VLD4:%.*]] = call { <1 x double>, <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld4r.v1f64.p0f64(double* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x double>, <1 x double>, <1 x double>, <1 x double> }*
+// CHECK:   store { <1 x double>, <1 x double>, <1 x double>, <1 x double> } [[VLD4]], { <1 x double>, <1 x double>, <1 x double>, <1 x double> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.float64x1x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP5:%.*]] = bitcast %struct.float64x1x4_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.float64x1x4_t, %struct.float64x1x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float64x1x4_t [[TMP6]]
 float64x1x4_t test_vld4_dup_f64(float64_t  *a) {
   return vld4_dup_f64(a);
 }
 
-// CHECK-LABEL: @test_vld4_dup_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X1X4_T]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
-// CHECK-NEXT:    [[VLD4:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4r.v1i64.p0i64(i64* [[TMP2]])
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP3]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast %struct.poly64x1x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP4]], i8* align 8 [[TMP5]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP6:%.*]] = load [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY64X1X4_T]] [[TMP6]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x1x4_t @test_vld4_dup_p64(i64* %a) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x4_t, align 8
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly64x1x4_t, align 8
+// CHECK:   [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP2:%.*]] = bitcast i8* [[TMP1]] to i64*
+// CHECK:   [[VLD4:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4r.v1i64.p0i64(i64* [[TMP2]])
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP3]]
+// CHECK:   [[TMP4:%.*]] = bitcast %struct.poly64x1x4_t* [[RETVAL]] to i8*
+// 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_vld4_dup_p64(poly64_t  *a) {
   return vld4_dup_p64(a);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <16 x i8> [[B:%.*]], i8 [[TMP0]], i32 15
-// CHECK-NEXT:    ret <16 x i8> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vld1q_lane_u8(i8* %a, <16 x i8> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <16 x i8> %b, i8 [[TMP0]], i32 15
+// CHECK:   ret <16 x i8> [[VLD1_LANE]]
 uint8x16_t test_vld1q_lane_u8(uint8_t  *a, uint8x16_t b) {
   return vld1q_lane_u8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[TMP4]], i32 7
-// CHECK-NEXT:    ret <8 x i16> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vld1q_lane_u16(i16* %a, <8 x i16> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP4:%.*]] = load i16, i16* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[TMP4]], i32 7
+// CHECK:   ret <8 x i16> [[VLD1_LANE]]
 uint16x8_t test_vld1q_lane_u16(uint16_t  *a, uint16x8_t b) {
   return vld1q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x i32> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i32> @test_vld1q_lane_u32(i32* %a, <4 x i32> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i32> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP4:%.*]] = load i32, i32* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP4]], i32 3
+// CHECK:   ret <4 x i32> [[VLD1_LANE]]
 uint32x4_t test_vld1q_lane_u32(uint32_t  *a, uint32x4_t b) {
   return vld1q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x i64> [[TMP2]], i64 [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x i64> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vld1q_lane_u64(i64* %a, <2 x i64> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP4:%.*]] = load i64, i64* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x i64> [[TMP2]], i64 [[TMP4]], i32 1
+// CHECK:   ret <2 x i64> [[VLD1_LANE]]
 uint64x2_t test_vld1q_lane_u64(uint64_t  *a, uint64x2_t b) {
   return vld1q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <16 x i8> [[B:%.*]], i8 [[TMP0]], i32 15
-// CHECK-NEXT:    ret <16 x i8> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vld1q_lane_s8(i8* %a, <16 x i8> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <16 x i8> %b, i8 [[TMP0]], i32 15
+// CHECK:   ret <16 x i8> [[VLD1_LANE]]
 int8x16_t test_vld1q_lane_s8(int8_t  *a, int8x16_t b) {
   return vld1q_lane_s8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[TMP4]], i32 7
-// CHECK-NEXT:    ret <8 x i16> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vld1q_lane_s16(i16* %a, <8 x i16> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP4:%.*]] = load i16, i16* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[TMP4]], i32 7
+// CHECK:   ret <8 x i16> [[VLD1_LANE]]
 int16x8_t test_vld1q_lane_s16(int16_t  *a, int16x8_t b) {
   return vld1q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x i32> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i32> @test_vld1q_lane_s32(i32* %a, <4 x i32> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i32> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP4:%.*]] = load i32, i32* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[TMP4]], i32 3
+// CHECK:   ret <4 x i32> [[VLD1_LANE]]
 int32x4_t test_vld1q_lane_s32(int32_t  *a, int32x4_t b) {
   return vld1q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x i64> [[TMP2]], i64 [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x i64> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vld1q_lane_s64(i64* %a, <2 x i64> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP4:%.*]] = load i64, i64* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x i64> [[TMP2]], i64 [[TMP4]], i32 1
+// CHECK:   ret <2 x i64> [[VLD1_LANE]]
 int64x2_t test_vld1q_lane_s64(int64_t  *a, int64x2_t b) {
   return vld1q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x half>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to half*
-// CHECK-NEXT:    [[TMP4:%.*]] = load half, half* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x half> [[TMP2]], half [[TMP4]], i32 7
-// CHECK-NEXT:    ret <8 x half> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x half> @test_vld1q_lane_f16(half* %a, <8 x half> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast half* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x half> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x half>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to half*
+// CHECK:   [[TMP4:%.*]] = load half, half* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x half> [[TMP2]], half [[TMP4]], i32 7
+// CHECK:   ret <8 x half> [[VLD1_LANE]]
 float16x8_t test_vld1q_lane_f16(float16_t  *a, float16x8_t b) {
   return vld1q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to float*
-// CHECK-NEXT:    [[TMP4:%.*]] = load float, float* [[TMP3]], align 4
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x float> [[TMP2]], float [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x float> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x float> @test_vld1q_lane_f32(float* %a, <4 x float> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast float* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x float> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to float*
+// CHECK:   [[TMP4:%.*]] = load float, float* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x float> [[TMP2]], float [[TMP4]], i32 3
+// CHECK:   ret <4 x float> [[VLD1_LANE]]
 float32x4_t test_vld1q_lane_f32(float32_t  *a, float32x4_t b) {
   return vld1q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to double*
-// CHECK-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x double> [[TMP2]], double [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x double> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x double> @test_vld1q_lane_f64(double* %a, <2 x double> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x double> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to double*
+// CHECK:   [[TMP4:%.*]] = load double, double* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x double> [[TMP2]], double [[TMP4]], i32 1
+// CHECK:   ret <2 x double> [[VLD1_LANE]]
 float64x2_t test_vld1q_lane_f64(float64_t  *a, float64x2_t b) {
   return vld1q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <16 x i8> [[B:%.*]], i8 [[TMP0]], i32 15
-// CHECK-NEXT:    ret <16 x i8> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vld1q_lane_p8(i8* %a, <16 x i8> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <16 x i8> %b, i8 [[TMP0]], i32 15
+// CHECK:   ret <16 x i8> [[VLD1_LANE]]
 poly8x16_t test_vld1q_lane_p8(poly8_t  *a, poly8x16_t b) {
   return vld1q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[TMP4]], i32 7
-// CHECK-NEXT:    ret <8 x i16> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vld1q_lane_p16(i16* %a, <8 x i16> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP4:%.*]] = load i16, i16* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x i16> [[TMP2]], i16 [[TMP4]], i32 7
+// CHECK:   ret <8 x i16> [[VLD1_LANE]]
 poly16x8_t test_vld1q_lane_p16(poly16_t  *a, poly16x8_t b) {
   return vld1q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x i64> [[TMP2]], i64 [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x i64> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vld1q_lane_p64(i64* %a, <2 x i64> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP4:%.*]] = load i64, i64* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x i64> [[TMP2]], i64 [[TMP4]], i32 1
+// CHECK:   ret <2 x i64> [[VLD1_LANE]]
 poly64x2_t test_vld1q_lane_p64(poly64_t  *a, poly64x2_t b) {
   return vld1q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x i8> [[B:%.*]], i8 [[TMP0]], i32 7
-// CHECK-NEXT:    ret <8 x i8> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i8> @test_vld1_lane_u8(i8* %a, <8 x i8> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x i8> %b, i8 [[TMP0]], i32 7
+// CHECK:   ret <8 x i8> [[VLD1_LANE]]
 uint8x8_t test_vld1_lane_u8(uint8_t  *a, uint8x8_t b) {
   return vld1_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x i16> [[TMP2]], i16 [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x i16> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i16> @test_vld1_lane_u16(i16* %a, <4 x i16> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP4:%.*]] = load i16, i16* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x i16> [[TMP2]], i16 [[TMP4]], i32 3
+// CHECK:   ret <4 x i16> [[VLD1_LANE]]
 uint16x4_t test_vld1_lane_u16(uint16_t  *a, uint16x4_t b) {
   return vld1_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x i32> [[TMP2]], i32 [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x i32> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i32> @test_vld1_lane_u32(i32* %a, <2 x i32> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i32> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP4:%.*]] = load i32, i32* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x i32> [[TMP2]], i32 [[TMP4]], i32 1
+// CHECK:   ret <2 x i32> [[VLD1_LANE]]
 uint32x2_t test_vld1_lane_u32(uint32_t  *a, uint32x2_t b) {
   return vld1_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <1 x i64> [[TMP2]], i64 [[TMP4]], i32 0
-// CHECK-NEXT:    ret <1 x i64> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x i64> @test_vld1_lane_u64(i64* %a, <1 x i64> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP4:%.*]] = load i64, i64* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <1 x i64> [[TMP2]], i64 [[TMP4]], i32 0
+// CHECK:   ret <1 x i64> [[VLD1_LANE]]
 uint64x1_t test_vld1_lane_u64(uint64_t  *a, uint64x1_t b) {
   return vld1_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld1_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x i8> [[B:%.*]], i8 [[TMP0]], i32 7
-// CHECK-NEXT:    ret <8 x i8> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i8> @test_vld1_lane_s8(i8* %a, <8 x i8> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x i8> %b, i8 [[TMP0]], i32 7
+// CHECK:   ret <8 x i8> [[VLD1_LANE]]
 int8x8_t test_vld1_lane_s8(int8_t  *a, int8x8_t b) {
   return vld1_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x i16> [[TMP2]], i16 [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x i16> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i16> @test_vld1_lane_s16(i16* %a, <4 x i16> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP4:%.*]] = load i16, i16* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x i16> [[TMP2]], i16 [[TMP4]], i32 3
+// CHECK:   ret <4 x i16> [[VLD1_LANE]]
 int16x4_t test_vld1_lane_s16(int16_t  *a, int16x4_t b) {
   return vld1_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x i32> [[TMP2]], i32 [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x i32> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i32> @test_vld1_lane_s32(i32* %a, <2 x i32> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i32> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   [[TMP4:%.*]] = load i32, i32* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x i32> [[TMP2]], i32 [[TMP4]], i32 1
+// CHECK:   ret <2 x i32> [[VLD1_LANE]]
 int32x2_t test_vld1_lane_s32(int32_t  *a, int32x2_t b) {
   return vld1_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <1 x i64> [[TMP2]], i64 [[TMP4]], i32 0
-// CHECK-NEXT:    ret <1 x i64> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x i64> @test_vld1_lane_s64(i64* %a, <1 x i64> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP4:%.*]] = load i64, i64* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <1 x i64> [[TMP2]], i64 [[TMP4]], i32 0
+// CHECK:   ret <1 x i64> [[VLD1_LANE]]
 int64x1_t test_vld1_lane_s64(int64_t  *a, int64x1_t b) {
   return vld1_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld1_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x half>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to half*
-// CHECK-NEXT:    [[TMP4:%.*]] = load half, half* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x half> [[TMP2]], half [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x half> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x half> @test_vld1_lane_f16(half* %a, <4 x half> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast half* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x half> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x half>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to half*
+// CHECK:   [[TMP4:%.*]] = load half, half* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x half> [[TMP2]], half [[TMP4]], i32 3
+// CHECK:   ret <4 x half> [[VLD1_LANE]]
 float16x4_t test_vld1_lane_f16(float16_t  *a, float16x4_t b) {
   return vld1_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to float*
-// CHECK-NEXT:    [[TMP4:%.*]] = load float, float* [[TMP3]], align 4
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP4]], i32 1
-// CHECK-NEXT:    ret <2 x float> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <2 x float> @test_vld1_lane_f32(float* %a, <2 x float> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast float* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x float> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to float*
+// CHECK:   [[TMP4:%.*]] = load float, float* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP4]], i32 1
+// CHECK:   ret <2 x float> [[VLD1_LANE]]
 float32x2_t test_vld1_lane_f32(float32_t  *a, float32x2_t b) {
   return vld1_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld1_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to double*
-// CHECK-NEXT:    [[TMP4:%.*]] = load double, double* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <1 x double> [[TMP2]], double [[TMP4]], i32 0
-// CHECK-NEXT:    ret <1 x double> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vld1_lane_f64(double* %a, <1 x double> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x double> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to double*
+// CHECK:   [[TMP4:%.*]] = load double, double* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <1 x double> [[TMP2]], double [[TMP4]], i32 0
+// CHECK:   ret <1 x double> [[VLD1_LANE]]
 float64x1_t test_vld1_lane_f64(float64_t  *a, float64x1_t b) {
   return vld1_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld1_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[A:%.*]], align 1
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <8 x i8> [[B:%.*]], i8 [[TMP0]], i32 7
-// CHECK-NEXT:    ret <8 x i8> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i8> @test_vld1_lane_p8(i8* %a, <8 x i8> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = load i8, i8* %a
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <8 x i8> %b, i8 [[TMP0]], i32 7
+// CHECK:   ret <8 x i8> [[VLD1_LANE]]
 poly8x8_t test_vld1_lane_p8(poly8_t  *a, poly8x8_t b) {
   return vld1_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld1_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 2
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <4 x i16> [[TMP2]], i16 [[TMP4]], i32 3
-// CHECK-NEXT:    ret <4 x i16> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i16> @test_vld1_lane_p16(i16* %a, <4 x i16> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   [[TMP4:%.*]] = load i16, i16* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <4 x i16> [[TMP2]], i16 [[TMP4]], i32 3
+// CHECK:   ret <4 x i16> [[VLD1_LANE]]
 poly16x4_t test_vld1_lane_p16(poly16_t  *a, poly16x4_t b) {
   return vld1_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld1_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-// CHECK-NEXT:    [[VLD1_LANE:%.*]] = insertelement <1 x i64> [[TMP2]], i64 [[TMP4]], i32 0
-// CHECK-NEXT:    ret <1 x i64> [[VLD1_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x i64> @test_vld1_lane_p64(i64* %a, <1 x i64> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// CHECK:   [[TMP3:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   [[TMP4:%.*]] = load i64, i64* [[TMP3]]
+// CHECK:   [[VLD1_LANE:%.*]] = insertelement <1 x i64> [[TMP2]], i64 [[TMP4]], i32 0
+// CHECK:   ret <1 x i64> [[VLD1_LANE]]
 poly64x1_t test_vld1_lane_p64(poly64_t  *a, poly64x1_t b) {
   return vld1_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT8X16X2_T:%.*]], align 16
-// CHECK-NEXT:    [[SRC:%.*]] = alloca [[STRUCT_INT8X16X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X16X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT8X16X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[SRC]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <16 x i8>] [[SRC_COERCE:%.*]], [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x16x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x16x2_t* [[SRC]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int8x16x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* [[PTR:%.*]])
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8> } [[VLD2_LANE]], { <16 x i8>, <16 x i8> }* [[TMP5]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast %struct.int8x16x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.int8x16x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP6]], i8* align 16 [[TMP7]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP8:%.*]] = load [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT8X16X2_T]] [[TMP8]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int8x16x2_t @test_vld2q_lane_s8(i8* %ptr, [2 x <16 x i8>] %src.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int8x16x2_t, align 16
+// CHECK:   [[SRC:%.*]] = alloca %struct.int8x16x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int8x16x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.int8x16x2_t, align 16
+// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x2_t, %struct.int8x16x2_t* [[SRC]], i32 0, i32 0
+// CHECK:   store [2 x <16 x i8>] [[SRC]].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* [[SRC]] 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 %struct.int8x16x2_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
+// CHECK:   [[VLD2_LANE:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* %ptr)
+// CHECK:   [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8> } [[VLD2_LANE]], { <16 x i8>, <16 x i8> }* [[TMP5]]
+// CHECK:   [[TMP6:%.*]] = bitcast %struct.int8x16x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.int8x16x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP6]], i8* align 16 [[TMP7]], i64 32, i1 false)
+// CHECK:   [[TMP8:%.*]] = load %struct.int8x16x2_t, %struct.int8x16x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int8x16x2_t [[TMP8]]
 int8x16x2_t test_vld2q_lane_s8(int8_t const * ptr, int8x16x2_t src) {
   return vld2q_lane_s8(ptr, src, 15);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT8X16X2_T:%.*]], align 16
-// CHECK-NEXT:    [[SRC:%.*]] = alloca [[STRUCT_UINT8X16X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X16X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT8X16X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[SRC]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <16 x i8>] [[SRC_COERCE:%.*]], [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x16x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x16x2_t* [[SRC]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint8x16x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* [[PTR:%.*]])
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8> } [[VLD2_LANE]], { <16 x i8>, <16 x i8> }* [[TMP5]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast %struct.uint8x16x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.uint8x16x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP6]], i8* align 16 [[TMP7]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP8:%.*]] = load [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT8X16X2_T]] [[TMP8]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint8x16x2_t @test_vld2q_lane_u8(i8* %ptr, [2 x <16 x i8>] %src.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint8x16x2_t, align 16
+// CHECK:   [[SRC:%.*]] = alloca %struct.uint8x16x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x16x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.uint8x16x2_t, align 16
+// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[SRC]], i32 0, i32 0
+// CHECK:   store [2 x <16 x i8>] [[SRC]].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* [[SRC]] 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 %struct.uint8x16x2_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
+// CHECK:   [[VLD2_LANE:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* %ptr)
+// CHECK:   [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8> } [[VLD2_LANE]], { <16 x i8>, <16 x i8> }* [[TMP5]]
+// CHECK:   [[TMP6:%.*]] = bitcast %struct.uint8x16x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.uint8x16x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP6]], i8* align 16 [[TMP7]], i64 32, i1 false)
+// CHECK:   [[TMP8:%.*]] = load %struct.uint8x16x2_t, %struct.uint8x16x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint8x16x2_t [[TMP8]]
 uint8x16x2_t test_vld2q_lane_u8(uint8_t const * ptr, uint8x16x2_t src) {
   return vld2q_lane_u8(ptr, src, 15);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY8X16X2_T:%.*]], align 16
-// CHECK-NEXT:    [[SRC:%.*]] = alloca [[STRUCT_POLY8X16X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X16X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY8X16X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[SRC]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <16 x i8>] [[SRC_COERCE:%.*]], [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x16x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x16x2_t* [[SRC]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly8x16x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* [[PTR:%.*]])
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8> } [[VLD2_LANE]], { <16 x i8>, <16 x i8> }* [[TMP5]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast %struct.poly8x16x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.poly8x16x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP6]], i8* align 16 [[TMP7]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP8:%.*]] = load [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY8X16X2_T]] [[TMP8]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly8x16x2_t @test_vld2q_lane_p8(i8* %ptr, [2 x <16 x i8>] %src.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly8x16x2_t, align 16
+// CHECK:   [[SRC:%.*]] = alloca %struct.poly8x16x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x16x2_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.poly8x16x2_t, align 16
+// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[SRC]], i32 0, i32 0
+// CHECK:   store [2 x <16 x i8>] [[SRC]].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* [[SRC]] 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 %struct.poly8x16x2_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
+// CHECK:   [[VLD2_LANE:%.*]] = call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* %ptr)
+// CHECK:   [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8> } [[VLD2_LANE]], { <16 x i8>, <16 x i8> }* [[TMP5]]
+// CHECK:   [[TMP6:%.*]] = bitcast %struct.poly8x16x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.poly8x16x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP6]], i8* align 16 [[TMP7]], i64 32, i1 false)
+// CHECK:   [[TMP8:%.*]] = load %struct.poly8x16x2_t, %struct.poly8x16x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly8x16x2_t [[TMP8]]
 poly8x16x2_t test_vld2q_lane_p8(poly8_t const * ptr, poly8x16x2_t src) {
   return vld2q_lane_p8(ptr, src, 15);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT8X16X3_T:%.*]], align 16
-// CHECK-NEXT:    [[SRC:%.*]] = alloca [[STRUCT_INT8X16X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X16X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT8X16X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[SRC]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <16 x i8>] [[SRC_COERCE:%.*]], [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x16x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x16x3_t* [[SRC]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int8x16x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld3lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* [[PTR:%.*]])
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8>, <16 x i8> } [[VLD3_LANE]], { <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP6]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.int8x16x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.int8x16x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP7]], i8* align 16 [[TMP8]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP9:%.*]] = load [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT8X16X3_T]] [[TMP9]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int8x16x3_t @test_vld3q_lane_s8(i8* %ptr, [3 x <16 x i8>] %src.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int8x16x3_t, align 16
+// CHECK:   [[SRC:%.*]] = alloca %struct.int8x16x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int8x16x3_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.int8x16x3_t, align 16
+// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x16x3_t, %struct.int8x16x3_t* [[SRC]], i32 0, i32 0
+// CHECK:   store [3 x <16 x i8>] [[SRC]].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* [[SRC]] 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 %struct.int8x16x3_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
+// CHECK:   [[VLD3_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld3lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* %ptr)
+// CHECK:   [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8>, <16 x i8> } [[VLD3_LANE]], { <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP6]]
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.int8x16x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.int8x16x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP7]], i8* align 16 [[TMP8]], i64 48, i1 false)
+// CHECK:   [[TMP9:%.*]] = load %struct.int8x16x3_t, %struct.int8x16x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int8x16x3_t [[TMP9]]
 int8x16x3_t test_vld3q_lane_s8(int8_t const * ptr, int8x16x3_t src) {
   return vld3q_lane_s8(ptr, src, 15);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT8X16X3_T:%.*]], align 16
-// CHECK-NEXT:    [[SRC:%.*]] = alloca [[STRUCT_UINT8X16X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X16X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT8X16X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[SRC]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <16 x i8>] [[SRC_COERCE:%.*]], [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x16x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x16x3_t* [[SRC]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint8x16x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld3lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* [[PTR:%.*]])
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8>, <16 x i8> } [[VLD3_LANE]], { <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP6]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.uint8x16x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.uint8x16x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP7]], i8* align 16 [[TMP8]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP9:%.*]] = load [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT8X16X3_T]] [[TMP9]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint8x16x3_t @test_vld3q_lane_u8(i8* %ptr, [3 x <16 x i8>] %src.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint8x16x3_t, align 16
+// CHECK:   [[SRC:%.*]] = alloca %struct.uint8x16x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x16x3_t, align 16
+// CHECK:   [[__RET:%.*]] = alloca %struct.uint8x16x3_t, align 16
+// CHECK:   [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[SRC]], i32 0, i32 0
+// CHECK:   store [3 x <16 x i8>] [[SRC]].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* [[SRC]] 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 %struct.uint8x16x3_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
+// CHECK:   [[VLD3_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld3lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* %ptr)
+// CHECK:   [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8>, <16 x i8> } [[VLD3_LANE]], { <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP6]]
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.uint8x16x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.uint8x16x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP7]], i8* align 16 [[TMP8]], i64 48, i1 false)
+// CHECK:   [[TMP9:%.*]] = load %struct.uint8x16x3_t, %struct.uint8x16x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint8x16x3_t [[TMP9]]
 uint8x16x3_t test_vld3q_lane_u8(uint8_t const * ptr, uint8x16x3_t src) {
   return vld3q_lane_u8(ptr, src, 15);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X8X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X8X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i16>] [[B_COERCE:%.*]], [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2lane.v8i16.p0i8(<8 x i16> [[TMP8]], <8 x i16> [[TMP9]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16> } [[VLD2_LANE]], { <8 x i16>, <8 x i16> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.uint16x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.uint16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT16X8X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint16x8x2_t @test_vld2q_lane_u16(i16* %a, [2 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint16x8x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint16x8x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x8x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint16x8x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2lane.v8i16.p0i8(<8 x i16> [[TMP8]], <8 x i16> [[TMP9]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16> } [[VLD2_LANE]], { <8 x i16>, <8 x i16> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.uint16x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.uint16x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint16x8x2_t [[TMP13]]
 uint16x8x2_t test_vld2q_lane_u16(uint16_t  *a, uint16x8x2_t b) {
   return vld2q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT32X4X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X4X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X4X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT32X4X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i32>] [[B_COERCE:%.*]], [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint32x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2lane.v4i32.p0i8(<4 x i32> [[TMP8]], <4 x i32> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32> }*
-// CHECK-NEXT:    store { <4 x i32>, <4 x i32> } [[VLD2_LANE]], { <4 x i32>, <4 x i32> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.uint32x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.uint32x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT32X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint32x4x2_t @test_vld2q_lane_u32(i32* %a, [2 x <4 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint32x4x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint32x4x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x4x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint32x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2lane.v4i32.p0i8(<4 x i32> [[TMP8]], <4 x i32> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32> }*
+// CHECK:   store { <4 x i32>, <4 x i32> } [[VLD2_LANE]], { <4 x i32>, <4 x i32> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.uint32x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.uint32x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint32x4x2_t [[TMP13]]
 uint32x4x2_t test_vld2q_lane_u32(uint32_t  *a, uint32x4x2_t b) {
   return vld2q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X2X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X2X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i64>] [[B_COERCE:%.*]], [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2lane.v2i64.p0i8(<2 x i64> [[TMP8]], <2 x i64> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64> } [[VLD2_LANE]], { <2 x i64>, <2 x i64> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.uint64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT64X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x2x2_t @test_vld2q_lane_u64(i64* %a, [2 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x2x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint64x2x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2lane.v2i64.p0i8(<2 x i64> [[TMP8]], <2 x i64> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2_LANE]], { <2 x i64>, <2 x i64> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.uint64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.uint64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.uint64x2x2_t, %struct.uint64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint64x2x2_t [[TMP13]]
 uint64x2x2_t test_vld2q_lane_u64(uint64_t  *a, uint64x2x2_t b) {
   return vld2q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X8X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X8X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i16>] [[B_COERCE:%.*]], [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2lane.v8i16.p0i8(<8 x i16> [[TMP8]], <8 x i16> [[TMP9]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16> } [[VLD2_LANE]], { <8 x i16>, <8 x i16> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.int16x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.int16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT16X8X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int16x8x2_t @test_vld2q_lane_s16(i16* %a, [2 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int16x8x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int16x8x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int16x8x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int16x8x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2lane.v8i16.p0i8(<8 x i16> [[TMP8]], <8 x i16> [[TMP9]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16> } [[VLD2_LANE]], { <8 x i16>, <8 x i16> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.int16x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.int16x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.int16x8x2_t, %struct.int16x8x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int16x8x2_t [[TMP13]]
 int16x8x2_t test_vld2q_lane_s16(int16_t  *a, int16x8x2_t b) {
   return vld2q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT32X4X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X4X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X4X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT32X4X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i32>] [[B_COERCE:%.*]], [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int32x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2lane.v4i32.p0i8(<4 x i32> [[TMP8]], <4 x i32> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32> }*
-// CHECK-NEXT:    store { <4 x i32>, <4 x i32> } [[VLD2_LANE]], { <4 x i32>, <4 x i32> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.int32x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.int32x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT32X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int32x4x2_t @test_vld2q_lane_s32(i32* %a, [2 x <4 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int32x4x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int32x4x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int32x4x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int32x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2lane.v4i32.p0i8(<4 x i32> [[TMP8]], <4 x i32> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32> }*
+// CHECK:   store { <4 x i32>, <4 x i32> } [[VLD2_LANE]], { <4 x i32>, <4 x i32> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.int32x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.int32x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.int32x4x2_t, %struct.int32x4x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int32x4x2_t [[TMP13]]
 int32x4x2_t test_vld2q_lane_s32(int32_t  *a, int32x4x2_t b) {
   return vld2q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X2X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X2X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i64>] [[B_COERCE:%.*]], [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2lane.v2i64.p0i8(<2 x i64> [[TMP8]], <2 x i64> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64> } [[VLD2_LANE]], { <2 x i64>, <2 x i64> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.int64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT64X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x2x2_t @test_vld2q_lane_s64(i64* %a, [2 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x2x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int64x2x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2lane.v2i64.p0i8(<2 x i64> [[TMP8]], <2 x i64> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2_LANE]], { <2 x i64>, <2 x i64> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.int64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.int64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.int64x2x2_t, %struct.int64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int64x2x2_t [[TMP13]]
 int64x2x2_t test_vld2q_lane_s64(int64_t  *a, int64x2x2_t b) {
   return vld2q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X8X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X8X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x half>] [[B_COERCE:%.*]], [2 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x half> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x half> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x half>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x half>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2lane.v8f16.p0i8(<8 x half> [[TMP8]], <8 x half> [[TMP9]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x half>, <8 x half> }*
-// CHECK-NEXT:    store { <8 x half>, <8 x half> } [[VLD2_LANE]], { <8 x half>, <8 x half> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.float16x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.float16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT16X8X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float16x8x2_t @test_vld2q_lane_f16(half* %a, [2 x <8 x half>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float16x8x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float16x8x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float16x8x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.float16x8x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x half> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x half> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x half>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x half>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2lane.v8f16.p0i8(<8 x half> [[TMP8]], <8 x half> [[TMP9]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x half>, <8 x half> }*
+// CHECK:   store { <8 x half>, <8 x half> } [[VLD2_LANE]], { <8 x half>, <8 x half> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.float16x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.float16x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.float16x8x2_t, %struct.float16x8x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float16x8x2_t [[TMP13]]
 float16x8x2_t test_vld2q_lane_f16(float16_t  *a, float16x8x2_t b) {
   return vld2q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT32X4X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X4X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X4X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT32X4X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x float>] [[B_COERCE:%.*]], [2 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float32x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x float> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x float> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x float>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x float>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x float>, <4 x float> } @llvm.aarch64.neon.ld2lane.v4f32.p0i8(<4 x float> [[TMP8]], <4 x float> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x float>, <4 x float> }*
-// CHECK-NEXT:    store { <4 x float>, <4 x float> } [[VLD2_LANE]], { <4 x float>, <4 x float> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.float32x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.float32x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT32X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float32x4x2_t @test_vld2q_lane_f32(float* %a, [2 x <4 x float>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float32x4x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float32x4x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float32x4x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.float32x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x float> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x float> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x float>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x float>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x float>, <4 x float> } @llvm.aarch64.neon.ld2lane.v4f32.p0i8(<4 x float> [[TMP8]], <4 x float> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x float>, <4 x float> }*
+// CHECK:   store { <4 x float>, <4 x float> } [[VLD2_LANE]], { <4 x float>, <4 x float> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.float32x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.float32x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.float32x4x2_t, %struct.float32x4x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float32x4x2_t [[TMP13]]
 float32x4x2_t test_vld2q_lane_f32(float32_t  *a, float32x4x2_t b) {
   return vld2q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x double>] [[B_COERCE:%.*]], [2 x <2 x double>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x double> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x double> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x double>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x double>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x double>, <2 x double> } @llvm.aarch64.neon.ld2lane.v2f64.p0i8(<2 x double> [[TMP8]], <2 x double> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x double>, <2 x double> }*
-// CHECK-NEXT:    store { <2 x double>, <2 x double> } [[VLD2_LANE]], { <2 x double>, <2 x double> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.float64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x2x2_t @test_vld2q_lane_f64(double* %a, [2 x <2 x double>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x2x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float64x2x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x2x2_t, %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x double> [[TMP4]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x2x2_t, %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP6:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x double> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x double>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x double>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x double>, <2 x double> } @llvm.aarch64.neon.ld2lane.v2f64.p0i8(<2 x double> [[TMP8]], <2 x double> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x double>, <2 x double> }*
+// CHECK:   store { <2 x double>, <2 x double> } [[VLD2_LANE]], { <2 x double>, <2 x double> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.float64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.float64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.float64x2x2_t, %struct.float64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float64x2x2_t [[TMP13]]
 float64x2x2_t test_vld2q_lane_f64(float64_t  *a, float64x2x2_t b) {
   return vld2q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X8X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X8X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i16>] [[B_COERCE:%.*]], [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2lane.v8i16.p0i8(<8 x i16> [[TMP8]], <8 x i16> [[TMP9]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16> } [[VLD2_LANE]], { <8 x i16>, <8 x i16> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.poly16x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.poly16x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY16X8X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly16x8x2_t @test_vld2q_lane_p16(i16* %a, [2 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly16x8x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly16x8x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x8x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.poly16x8x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld2lane.v8i16.p0i8(<8 x i16> [[TMP8]], <8 x i16> [[TMP9]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16> } [[VLD2_LANE]], { <8 x i16>, <8 x i16> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.poly16x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.poly16x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly16x8x2_t [[TMP13]]
 poly16x8x2_t test_vld2q_lane_p16(poly16_t  *a, poly16x8x2_t b) {
   return vld2q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X2X2_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X2X2_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i64>] [[B_COERCE:%.*]], [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2lane.v2i64.p0i8(<2 x i64> [[TMP8]], <2 x i64> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64> } [[VLD2_LANE]], { <2 x i64>, <2 x i64> }* [[TMP10]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.poly64x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY64X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x2x2_t @test_vld2q_lane_p64(i64* %a, [2 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x2_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly64x2x2_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x2_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld2lane.v2i64.p0i8(<2 x i64> [[TMP8]], <2 x i64> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64> } [[VLD2_LANE]], { <2 x i64>, <2 x i64> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.poly64x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.poly64x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP11]], i8* align 16 [[TMP12]], i64 32, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.poly64x2x2_t, %struct.poly64x2x2_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly64x2x2_t [[TMP13]]
 poly64x2x2_t test_vld2q_lane_p64(poly64_t  *a, poly64x2x2_t b) {
   return vld2q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT8X8X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X8X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X8X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT8X8X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i8>] [[B_COERCE:%.*]], [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint8x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld2lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8> } [[VLD2_LANE]], { <8 x i8>, <8 x i8> }* [[TMP5]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast %struct.uint8x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.uint8x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP8:%.*]] = load [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT8X8X2_T]] [[TMP8]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint8x8x2_t @test_vld2_lane_u8(i8* %a, [2 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint8x8x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint8x8x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x8x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.uint8x8x2_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld2lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* %a)
+// CHECK:   [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8> } [[VLD2_LANE]], { <8 x i8>, <8 x i8> }* [[TMP5]]
+// CHECK:   [[TMP6:%.*]] = bitcast %struct.uint8x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.uint8x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false)
+// CHECK:   [[TMP8:%.*]] = load %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint8x8x2_t [[TMP8]]
 uint8x8x2_t test_vld2_lane_u8(uint8_t  *a, uint8x8x2_t b) {
   return vld2_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X4X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X4X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i16>] [[B_COERCE:%.*]], [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld2lane.v4i16.p0i8(<4 x i16> [[TMP8]], <4 x i16> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16> } [[VLD2_LANE]], { <4 x i16>, <4 x i16> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.uint16x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.uint16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT16X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint16x4x2_t @test_vld2_lane_u16(i16* %a, [2 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint16x4x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint16x4x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x4x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint16x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld2lane.v4i16.p0i8(<4 x i16> [[TMP8]], <4 x i16> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16> } [[VLD2_LANE]], { <4 x i16>, <4 x i16> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.uint16x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.uint16x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint16x4x2_t [[TMP13]]
 uint16x4x2_t test_vld2_lane_u16(uint16_t  *a, uint16x4x2_t b) {
   return vld2_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT32X2X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X2X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X2X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT32X2X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i32>] [[B_COERCE:%.*]], [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint32x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2lane.v2i32.p0i8(<2 x i32> [[TMP8]], <2 x i32> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32> }*
-// CHECK-NEXT:    store { <2 x i32>, <2 x i32> } [[VLD2_LANE]], { <2 x i32>, <2 x i32> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.uint32x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.uint32x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT32X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint32x2x2_t @test_vld2_lane_u32(i32* %a, [2 x <2 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint32x2x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint32x2x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x2x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint32x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2lane.v2i32.p0i8(<2 x i32> [[TMP8]], <2 x i32> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32> }*
+// CHECK:   store { <2 x i32>, <2 x i32> } [[VLD2_LANE]], { <2 x i32>, <2 x i32> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.uint32x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.uint32x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint32x2x2_t [[TMP13]]
 uint32x2x2_t test_vld2_lane_u32(uint32_t  *a, uint32x2x2_t b) {
   return vld2_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X1X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X1X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x i64>] [[B_COERCE:%.*]], [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2lane.v1i64.p0i8(<1 x i64> [[TMP8]], <1 x i64> [[TMP9]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64> } [[VLD2_LANE]], { <1 x i64>, <1 x i64> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.uint64x1x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.uint64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT64X1X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x1x2_t @test_vld2_lane_u64(i64* %a, [2 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x1x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint64x1x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x1x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint64x1x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2lane.v1i64.p0i8(<1 x i64> [[TMP8]], <1 x i64> [[TMP9]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64> } [[VLD2_LANE]], { <1 x i64>, <1 x i64> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.uint64x1x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.uint64x1x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.uint64x1x2_t, %struct.uint64x1x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint64x1x2_t [[TMP13]]
 uint64x1x2_t test_vld2_lane_u64(uint64_t  *a, uint64x1x2_t b) {
   return vld2_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld2_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT8X8X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X8X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X8X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT8X8X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i8>] [[B_COERCE:%.*]], [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int8x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld2lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8> } [[VLD2_LANE]], { <8 x i8>, <8 x i8> }* [[TMP5]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast %struct.int8x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.int8x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP8:%.*]] = load [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT8X8X2_T]] [[TMP8]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int8x8x2_t @test_vld2_lane_s8(i8* %a, [2 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int8x8x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int8x8x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int8x8x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.int8x8x2_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld2lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* %a)
+// CHECK:   [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8> } [[VLD2_LANE]], { <8 x i8>, <8 x i8> }* [[TMP5]]
+// CHECK:   [[TMP6:%.*]] = bitcast %struct.int8x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.int8x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false)
+// CHECK:   [[TMP8:%.*]] = load %struct.int8x8x2_t, %struct.int8x8x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int8x8x2_t [[TMP8]]
 int8x8x2_t test_vld2_lane_s8(int8_t  *a, int8x8x2_t b) {
   return vld2_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X4X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X4X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i16>] [[B_COERCE:%.*]], [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld2lane.v4i16.p0i8(<4 x i16> [[TMP8]], <4 x i16> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16> } [[VLD2_LANE]], { <4 x i16>, <4 x i16> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.int16x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.int16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT16X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int16x4x2_t @test_vld2_lane_s16(i16* %a, [2 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int16x4x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int16x4x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int16x4x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int16x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld2lane.v4i16.p0i8(<4 x i16> [[TMP8]], <4 x i16> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16> } [[VLD2_LANE]], { <4 x i16>, <4 x i16> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.int16x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.int16x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.int16x4x2_t, %struct.int16x4x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int16x4x2_t [[TMP13]]
 int16x4x2_t test_vld2_lane_s16(int16_t  *a, int16x4x2_t b) {
   return vld2_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT32X2X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X2X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X2X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT32X2X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i32>] [[B_COERCE:%.*]], [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int32x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2lane.v2i32.p0i8(<2 x i32> [[TMP8]], <2 x i32> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32> }*
-// CHECK-NEXT:    store { <2 x i32>, <2 x i32> } [[VLD2_LANE]], { <2 x i32>, <2 x i32> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.int32x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.int32x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT32X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int32x2x2_t @test_vld2_lane_s32(i32* %a, [2 x <2 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int32x2x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int32x2x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int32x2x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int32x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2lane.v2i32.p0i8(<2 x i32> [[TMP8]], <2 x i32> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32> }*
+// CHECK:   store { <2 x i32>, <2 x i32> } [[VLD2_LANE]], { <2 x i32>, <2 x i32> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.int32x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.int32x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.int32x2x2_t, %struct.int32x2x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int32x2x2_t [[TMP13]]
 int32x2x2_t test_vld2_lane_s32(int32_t  *a, int32x2x2_t b) {
   return vld2_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X1X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X1X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x i64>] [[B_COERCE:%.*]], [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2lane.v1i64.p0i8(<1 x i64> [[TMP8]], <1 x i64> [[TMP9]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64> } [[VLD2_LANE]], { <1 x i64>, <1 x i64> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.int64x1x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.int64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT64X1X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x1x2_t @test_vld2_lane_s64(i64* %a, [2 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x1x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int64x1x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int64x1x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int64x1x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2lane.v1i64.p0i8(<1 x i64> [[TMP8]], <1 x i64> [[TMP9]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64> } [[VLD2_LANE]], { <1 x i64>, <1 x i64> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.int64x1x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.int64x1x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.int64x1x2_t, %struct.int64x1x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int64x1x2_t [[TMP13]]
 int64x1x2_t test_vld2_lane_s64(int64_t  *a, int64x1x2_t b) {
   return vld2_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld2_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X4X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X4X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x half>] [[B_COERCE:%.*]], [2 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x half> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x half> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x half>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x half>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2lane.v4f16.p0i8(<4 x half> [[TMP8]], <4 x half> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x half>, <4 x half> }*
-// CHECK-NEXT:    store { <4 x half>, <4 x half> } [[VLD2_LANE]], { <4 x half>, <4 x half> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.float16x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.float16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT16X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float16x4x2_t @test_vld2_lane_f16(half* %a, [2 x <4 x half>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float16x4x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float16x4x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float16x4x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.float16x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x half> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x half> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x half>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x half>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2lane.v4f16.p0i8(<4 x half> [[TMP8]], <4 x half> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x half>, <4 x half> }*
+// CHECK:   store { <4 x half>, <4 x half> } [[VLD2_LANE]], { <4 x half>, <4 x half> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.float16x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.float16x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.float16x4x2_t, %struct.float16x4x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float16x4x2_t [[TMP13]]
 float16x4x2_t test_vld2_lane_f16(float16_t  *a, float16x4x2_t b) {
   return vld2_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT32X2X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X2X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X2X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT32X2X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x float>] [[B_COERCE:%.*]], [2 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float32x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x float> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x float>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x float>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <2 x float>, <2 x float> } @llvm.aarch64.neon.ld2lane.v2f32.p0i8(<2 x float> [[TMP8]], <2 x float> [[TMP9]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x float>, <2 x float> }*
-// CHECK-NEXT:    store { <2 x float>, <2 x float> } [[VLD2_LANE]], { <2 x float>, <2 x float> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.float32x2x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.float32x2x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT32X2X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float32x2x2_t @test_vld2_lane_f32(float* %a, [2 x <2 x float>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float32x2x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float32x2x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float32x2x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.float32x2x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x float> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x float>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x float>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <2 x float>, <2 x float> } @llvm.aarch64.neon.ld2lane.v2f32.p0i8(<2 x float> [[TMP8]], <2 x float> [[TMP9]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <2 x float>, <2 x float> }*
+// CHECK:   store { <2 x float>, <2 x float> } [[VLD2_LANE]], { <2 x float>, <2 x float> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.float32x2x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.float32x2x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.float32x2x2_t, %struct.float32x2x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float32x2x2_t [[TMP13]]
 float32x2x2_t test_vld2_lane_f32(float32_t  *a, float32x2x2_t b) {
   return vld2_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld2_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x double>] [[B_COERCE:%.*]], [2 x <1 x double>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x double> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x double> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x double>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x double>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <1 x double>, <1 x double> } @llvm.aarch64.neon.ld2lane.v1f64.p0i8(<1 x double> [[TMP8]], <1 x double> [[TMP9]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x double>, <1 x double> }*
-// CHECK-NEXT:    store { <1 x double>, <1 x double> } [[VLD2_LANE]], { <1 x double>, <1 x double> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.float64x1x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X1X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x1x2_t @test_vld2_lane_f64(double* %a, [2 x <1 x double>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x1x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float64x1x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float64x1x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x1x2_t, %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP4:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x double> [[TMP4]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x1x2_t, %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP6:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x double> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x double>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x double>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <1 x double>, <1 x double> } @llvm.aarch64.neon.ld2lane.v1f64.p0i8(<1 x double> [[TMP8]], <1 x double> [[TMP9]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x double>, <1 x double> }*
+// CHECK:   store { <1 x double>, <1 x double> } [[VLD2_LANE]], { <1 x double>, <1 x double> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.float64x1x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.float64x1x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.float64x1x2_t, %struct.float64x1x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float64x1x2_t [[TMP13]]
 float64x1x2_t test_vld2_lane_f64(float64_t  *a, float64x1x2_t b) {
   return vld2_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld2_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY8X8X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X8X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X8X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY8X8X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i8>] [[B_COERCE:%.*]], [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly8x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld2lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8> } [[VLD2_LANE]], { <8 x i8>, <8 x i8> }* [[TMP5]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast %struct.poly8x8x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.poly8x8x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP8:%.*]] = load [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY8X8X2_T]] [[TMP8]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly8x8x2_t @test_vld2_lane_p8(i8* %a, [2 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly8x8x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly8x8x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x8x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly8x8x2_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
+// CHECK:   [[VLD2_LANE:%.*]] = call { <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld2lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* %a)
+// CHECK:   [[TMP5:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8> } [[VLD2_LANE]], { <8 x i8>, <8 x i8> }* [[TMP5]]
+// CHECK:   [[TMP6:%.*]] = bitcast %struct.poly8x8x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.poly8x8x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false)
+// CHECK:   [[TMP8:%.*]] = load %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly8x8x2_t [[TMP8]]
 poly8x8x2_t test_vld2_lane_p8(poly8_t  *a, poly8x8x2_t b) {
   return vld2_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld2_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X4X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X4X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i16>] [[B_COERCE:%.*]], [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld2lane.v4i16.p0i8(<4 x i16> [[TMP8]], <4 x i16> [[TMP9]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16> } [[VLD2_LANE]], { <4 x i16>, <4 x i16> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.poly16x4x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.poly16x4x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY16X4X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly16x4x2_t @test_vld2_lane_p16(i16* %a, [2 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly16x4x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly16x4x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x4x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.poly16x4x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld2lane.v4i16.p0i8(<4 x i16> [[TMP8]], <4 x i16> [[TMP9]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16> } [[VLD2_LANE]], { <4 x i16>, <4 x i16> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.poly16x4x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.poly16x4x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly16x4x2_t [[TMP13]]
 poly16x4x2_t test_vld2_lane_p16(poly16_t  *a, poly16x4x2_t b) {
   return vld2_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld2_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X1X2_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X1X2_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x i64>] [[B_COERCE:%.*]], [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[VLD2_LANE:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2lane.v1i64.p0i8(<1 x i64> [[TMP8]], <1 x i64> [[TMP9]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64> } [[VLD2_LANE]], { <1 x i64>, <1 x i64> }* [[TMP10]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast %struct.poly64x1x2_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP13:%.*]] = load [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY64X1X2_T]] [[TMP13]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x1x2_t @test_vld2_lane_p64(i64* %a, [2 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x2_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly64x1x2_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x2_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[VLD2_LANE:%.*]] = call { <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld2lane.v1i64.p0i8(<1 x i64> [[TMP8]], <1 x i64> [[TMP9]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP10:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64> } [[VLD2_LANE]], { <1 x i64>, <1 x i64> }* [[TMP10]]
+// CHECK:   [[TMP11:%.*]] = bitcast %struct.poly64x1x2_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP12:%.*]] = bitcast %struct.poly64x1x2_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP11]], i8* align 8 [[TMP12]], i64 16, i1 false)
+// CHECK:   [[TMP13:%.*]] = load %struct.poly64x1x2_t, %struct.poly64x1x2_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly64x1x2_t [[TMP13]]
 poly64x1x2_t test_vld2_lane_p64(poly64_t  *a, poly64x1x2_t b) {
   return vld2_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X8X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X8X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i16>] [[B_COERCE:%.*]], [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld3lane.v8i16.p0i8(<8 x i16> [[TMP10]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16>, <8 x i16> } [[VLD3_LANE]], { <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.uint16x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.uint16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT16X8X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint16x8x3_t @test_vld3q_lane_u16(i16* %a, [3 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint16x8x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint16x8x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x8x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint16x8x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld3lane.v8i16.p0i8(<8 x i16> [[TMP10]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16>, <8 x i16> } [[VLD3_LANE]], { <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.uint16x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.uint16x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint16x8x3_t [[TMP16]]
 uint16x8x3_t test_vld3q_lane_u16(uint16_t  *a, uint16x8x3_t b) {
   return vld3q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT32X4X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X4X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X4X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT32X4X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i32>] [[B_COERCE:%.*]], [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint32x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3lane.v4i32.p0i8(<4 x i32> [[TMP10]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32> }*
-// CHECK-NEXT:    store { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3_LANE]], { <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.uint32x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.uint32x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT32X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint32x4x3_t @test_vld3q_lane_u32(i32* %a, [3 x <4 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint32x4x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint32x4x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x4x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint32x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3lane.v4i32.p0i8(<4 x i32> [[TMP10]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32> }*
+// CHECK:   store { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3_LANE]], { <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.uint32x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.uint32x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint32x4x3_t [[TMP16]]
 uint32x4x3_t test_vld3q_lane_u32(uint32_t  *a, uint32x4x3_t b) {
   return vld3q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X2X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X2X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i64>] [[B_COERCE:%.*]], [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3lane.v2i64.p0i8(<2 x i64> [[TMP10]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3_LANE]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.uint64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT64X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x2x3_t @test_vld3q_lane_u64(i64* %a, [3 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x2x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint64x2x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3lane.v2i64.p0i8(<2 x i64> [[TMP10]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3_LANE]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.uint64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.uint64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.uint64x2x3_t, %struct.uint64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint64x2x3_t [[TMP16]]
 uint64x2x3_t test_vld3q_lane_u64(uint64_t  *a, uint64x2x3_t b) {
   return vld3q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X8X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X8X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i16>] [[B_COERCE:%.*]], [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld3lane.v8i16.p0i8(<8 x i16> [[TMP10]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16>, <8 x i16> } [[VLD3_LANE]], { <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.int16x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.int16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT16X8X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int16x8x3_t @test_vld3q_lane_s16(i16* %a, [3 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int16x8x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int16x8x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int16x8x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int16x8x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld3lane.v8i16.p0i8(<8 x i16> [[TMP10]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16>, <8 x i16> } [[VLD3_LANE]], { <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.int16x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.int16x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.int16x8x3_t, %struct.int16x8x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int16x8x3_t [[TMP16]]
 int16x8x3_t test_vld3q_lane_s16(int16_t  *a, int16x8x3_t b) {
   return vld3q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT32X4X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X4X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X4X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT32X4X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i32>] [[B_COERCE:%.*]], [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int32x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3lane.v4i32.p0i8(<4 x i32> [[TMP10]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32> }*
-// CHECK-NEXT:    store { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3_LANE]], { <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.int32x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.int32x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT32X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int32x4x3_t @test_vld3q_lane_s32(i32* %a, [3 x <4 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int32x4x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int32x4x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int32x4x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int32x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld3lane.v4i32.p0i8(<4 x i32> [[TMP10]], <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32> }*
+// CHECK:   store { <4 x i32>, <4 x i32>, <4 x i32> } [[VLD3_LANE]], { <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.int32x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.int32x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.int32x4x3_t, %struct.int32x4x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int32x4x3_t [[TMP16]]
 int32x4x3_t test_vld3q_lane_s32(int32_t  *a, int32x4x3_t b) {
   return vld3q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X2X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X2X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i64>] [[B_COERCE:%.*]], [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3lane.v2i64.p0i8(<2 x i64> [[TMP10]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3_LANE]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.int64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT64X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x2x3_t @test_vld3q_lane_s64(i64* %a, [3 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x2x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int64x2x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3lane.v2i64.p0i8(<2 x i64> [[TMP10]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3_LANE]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.int64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.int64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.int64x2x3_t, %struct.int64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int64x2x3_t [[TMP16]]
 int64x2x3_t test_vld3q_lane_s64(int64_t  *a, int64x2x3_t b) {
   return vld3q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X8X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X8X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x half>] [[B_COERCE:%.*]], [3 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x half> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x half> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x half> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x half>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x half>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x half>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3lane.v8f16.p0i8(<8 x half> [[TMP10]], <8 x half> [[TMP11]], <8 x half> [[TMP12]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x half>, <8 x half>, <8 x half> }*
-// CHECK-NEXT:    store { <8 x half>, <8 x half>, <8 x half> } [[VLD3_LANE]], { <8 x half>, <8 x half>, <8 x half> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.float16x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.float16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT16X8X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float16x8x3_t @test_vld3q_lane_f16(half* %a, [3 x <8 x half>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float16x8x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float16x8x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float16x8x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.float16x8x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x half> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x half> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x half> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x half>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x half>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x half>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3lane.v8f16.p0i8(<8 x half> [[TMP10]], <8 x half> [[TMP11]], <8 x half> [[TMP12]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x half>, <8 x half>, <8 x half> }*
+// CHECK:   store { <8 x half>, <8 x half>, <8 x half> } [[VLD3_LANE]], { <8 x half>, <8 x half>, <8 x half> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.float16x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.float16x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.float16x8x3_t, %struct.float16x8x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float16x8x3_t [[TMP16]]
 float16x8x3_t test_vld3q_lane_f16(float16_t  *a, float16x8x3_t b) {
   return vld3q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT32X4X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X4X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X4X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT32X4X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x float>] [[B_COERCE:%.*]], [3 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float32x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x float> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x float> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x float> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x float>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x float>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x float>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld3lane.v4f32.p0i8(<4 x float> [[TMP10]], <4 x float> [[TMP11]], <4 x float> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x float>, <4 x float>, <4 x float> }*
-// CHECK-NEXT:    store { <4 x float>, <4 x float>, <4 x float> } [[VLD3_LANE]], { <4 x float>, <4 x float>, <4 x float> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.float32x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.float32x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT32X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float32x4x3_t @test_vld3q_lane_f32(float* %a, [3 x <4 x float>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float32x4x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float32x4x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float32x4x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.float32x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x float> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x float> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x float> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x float>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x float>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x float>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld3lane.v4f32.p0i8(<4 x float> [[TMP10]], <4 x float> [[TMP11]], <4 x float> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x float>, <4 x float>, <4 x float> }*
+// CHECK:   store { <4 x float>, <4 x float>, <4 x float> } [[VLD3_LANE]], { <4 x float>, <4 x float>, <4 x float> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.float32x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.float32x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.float32x4x3_t, %struct.float32x4x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float32x4x3_t [[TMP16]]
 float32x4x3_t test_vld3q_lane_f32(float32_t  *a, float32x4x3_t b) {
   return vld3q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x double>] [[B_COERCE:%.*]], [3 x <2 x double>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x double> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x double> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x double> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x double>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x double>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x double>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld3lane.v2f64.p0i8(<2 x double> [[TMP10]], <2 x double> [[TMP11]], <2 x double> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x double>, <2 x double>, <2 x double> }*
-// CHECK-NEXT:    store { <2 x double>, <2 x double>, <2 x double> } [[VLD3_LANE]], { <2 x double>, <2 x double>, <2 x double> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.float64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x2x3_t @test_vld3q_lane_f64(double* %a, [3 x <2 x double>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x2x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float64x2x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x double> [[TMP4]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP6:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x double> [[TMP6]] to <16 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP8:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x double> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x double>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x double>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x double>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld3lane.v2f64.p0i8(<2 x double> [[TMP10]], <2 x double> [[TMP11]], <2 x double> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x double>, <2 x double>, <2 x double> }*
+// CHECK:   store { <2 x double>, <2 x double>, <2 x double> } [[VLD3_LANE]], { <2 x double>, <2 x double>, <2 x double> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.float64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.float64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.float64x2x3_t, %struct.float64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float64x2x3_t [[TMP16]]
 float64x2x3_t test_vld3q_lane_f64(float64_t  *a, float64x2x3_t b) {
   return vld3q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY8X16X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X16X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X16X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY8X16X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <16 x i8>] [[B_COERCE:%.*]], [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x16x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x16x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly8x16x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld3lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8>, <16 x i8> } [[VLD3_LANE]], { <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP6]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.poly8x16x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.poly8x16x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP7]], i8* align 16 [[TMP8]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP9:%.*]] = load [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY8X16X3_T]] [[TMP9]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly8x16x3_t @test_vld3q_lane_p8(i8* %a, [3 x <16 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly8x16x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly8x16x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x16x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly8x16x3_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
+// CHECK:   [[VLD3_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld3lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* %a)
+// CHECK:   [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8>, <16 x i8> } [[VLD3_LANE]], { <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP6]]
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.poly8x16x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.poly8x16x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP7]], i8* align 16 [[TMP8]], i64 48, i1 false)
+// CHECK:   [[TMP9:%.*]] = load %struct.poly8x16x3_t, %struct.poly8x16x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly8x16x3_t [[TMP9]]
 poly8x16x3_t test_vld3q_lane_p8(poly8_t  *a, poly8x16x3_t b) {
   return vld3q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X8X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X8X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i16>] [[B_COERCE:%.*]], [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld3lane.v8i16.p0i8(<8 x i16> [[TMP10]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16>, <8 x i16> } [[VLD3_LANE]], { <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.poly16x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.poly16x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY16X8X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly16x8x3_t @test_vld3q_lane_p16(i16* %a, [3 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly16x8x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly16x8x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x8x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.poly16x8x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld3lane.v8i16.p0i8(<8 x i16> [[TMP10]], <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16>, <8 x i16> } [[VLD3_LANE]], { <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.poly16x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.poly16x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly16x8x3_t [[TMP16]]
 poly16x8x3_t test_vld3q_lane_p16(poly16_t  *a, poly16x8x3_t b) {
   return vld3q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X2X3_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X2X3_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i64>] [[B_COERCE:%.*]], [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3lane.v2i64.p0i8(<2 x i64> [[TMP10]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3_LANE]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP13]], align 16
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.poly64x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY64X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x2x3_t @test_vld3q_lane_p64(i64* %a, [3 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x3_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly64x2x3_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x3_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_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:   [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld3lane.v2i64.p0i8(<2 x i64> [[TMP10]], <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64> } [[VLD3_LANE]], { <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.poly64x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.poly64x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP14]], i8* align 16 [[TMP15]], i64 48, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.poly64x2x3_t, %struct.poly64x2x3_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly64x2x3_t [[TMP16]]
 poly64x2x3_t test_vld3q_lane_p64(poly64_t  *a, poly64x2x3_t b) {
   return vld3q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT8X8X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X8X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X8X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT8X8X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i8>] [[B_COERCE:%.*]], [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint8x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld3lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8>, <8 x i8> } [[VLD3_LANE]], { <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP6]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.uint8x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.uint8x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 [[TMP8]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP9:%.*]] = load [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT8X8X3_T]] [[TMP9]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint8x8x3_t @test_vld3_lane_u8(i8* %a, [3 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint8x8x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint8x8x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x8x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.uint8x8x3_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld3lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* %a)
+// CHECK:   [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8>, <8 x i8> } [[VLD3_LANE]], { <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP6]]
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.uint8x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.uint8x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 [[TMP8]], i64 24, i1 false)
+// CHECK:   [[TMP9:%.*]] = load %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint8x8x3_t [[TMP9]]
 uint8x8x3_t test_vld3_lane_u8(uint8_t  *a, uint8x8x3_t b) {
   return vld3_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X4X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X4X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i16>] [[B_COERCE:%.*]], [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3lane.v4i16.p0i8(<4 x i16> [[TMP10]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16>, <4 x i16> } [[VLD3_LANE]], { <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.uint16x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.uint16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT16X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint16x4x3_t @test_vld3_lane_u16(i16* %a, [3 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint16x4x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint16x4x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x4x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint16x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3lane.v4i16.p0i8(<4 x i16> [[TMP10]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16>, <4 x i16> } [[VLD3_LANE]], { <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.uint16x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.uint16x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint16x4x3_t [[TMP16]]
 uint16x4x3_t test_vld3_lane_u16(uint16_t  *a, uint16x4x3_t b) {
   return vld3_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT32X2X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X2X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X2X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT32X2X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i32>] [[B_COERCE:%.*]], [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint32x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld3lane.v2i32.p0i8(<2 x i32> [[TMP10]], <2 x i32> [[TMP11]], <2 x i32> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32> }*
-// CHECK-NEXT:    store { <2 x i32>, <2 x i32>, <2 x i32> } [[VLD3_LANE]], { <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.uint32x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.uint32x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT32X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint32x2x3_t @test_vld3_lane_u32(i32* %a, [3 x <2 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint32x2x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint32x2x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x2x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint32x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld3lane.v2i32.p0i8(<2 x i32> [[TMP10]], <2 x i32> [[TMP11]], <2 x i32> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32> }*
+// CHECK:   store { <2 x i32>, <2 x i32>, <2 x i32> } [[VLD3_LANE]], { <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.uint32x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.uint32x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint32x2x3_t [[TMP16]]
 uint32x2x3_t test_vld3_lane_u32(uint32_t  *a, uint32x2x3_t b) {
   return vld3_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X1X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X1X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x i64>] [[B_COERCE:%.*]], [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3lane.v1i64.p0i8(<1 x i64> [[TMP10]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3_LANE]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.uint64x1x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.uint64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT64X1X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x1x3_t @test_vld3_lane_u64(i64* %a, [3 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x1x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint64x1x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x1x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint64x1x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3lane.v1i64.p0i8(<1 x i64> [[TMP10]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3_LANE]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.uint64x1x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.uint64x1x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.uint64x1x3_t, %struct.uint64x1x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint64x1x3_t [[TMP16]]
 uint64x1x3_t test_vld3_lane_u64(uint64_t  *a, uint64x1x3_t b) {
   return vld3_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld3_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT8X8X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X8X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X8X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT8X8X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i8>] [[B_COERCE:%.*]], [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int8x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld3lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8>, <8 x i8> } [[VLD3_LANE]], { <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP6]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.int8x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.int8x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 [[TMP8]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP9:%.*]] = load [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT8X8X3_T]] [[TMP9]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int8x8x3_t @test_vld3_lane_s8(i8* %a, [3 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int8x8x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int8x8x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int8x8x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.int8x8x3_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld3lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* %a)
+// CHECK:   [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8>, <8 x i8> } [[VLD3_LANE]], { <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP6]]
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.int8x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.int8x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 [[TMP8]], i64 24, i1 false)
+// CHECK:   [[TMP9:%.*]] = load %struct.int8x8x3_t, %struct.int8x8x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int8x8x3_t [[TMP9]]
 int8x8x3_t test_vld3_lane_s8(int8_t  *a, int8x8x3_t b) {
   return vld3_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X4X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X4X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i16>] [[B_COERCE:%.*]], [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3lane.v4i16.p0i8(<4 x i16> [[TMP10]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16>, <4 x i16> } [[VLD3_LANE]], { <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.int16x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.int16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT16X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int16x4x3_t @test_vld3_lane_s16(i16* %a, [3 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int16x4x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int16x4x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int16x4x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int16x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3lane.v4i16.p0i8(<4 x i16> [[TMP10]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16>, <4 x i16> } [[VLD3_LANE]], { <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.int16x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.int16x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.int16x4x3_t, %struct.int16x4x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int16x4x3_t [[TMP16]]
 int16x4x3_t test_vld3_lane_s16(int16_t  *a, int16x4x3_t b) {
   return vld3_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT32X2X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X2X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X2X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT32X2X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i32>] [[B_COERCE:%.*]], [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int32x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld3lane.v2i32.p0i8(<2 x i32> [[TMP10]], <2 x i32> [[TMP11]], <2 x i32> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32> }*
-// CHECK-NEXT:    store { <2 x i32>, <2 x i32>, <2 x i32> } [[VLD3_LANE]], { <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.int32x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.int32x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT32X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int32x2x3_t @test_vld3_lane_s32(i32* %a, [3 x <2 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int32x2x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int32x2x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int32x2x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int32x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld3lane.v2i32.p0i8(<2 x i32> [[TMP10]], <2 x i32> [[TMP11]], <2 x i32> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32> }*
+// CHECK:   store { <2 x i32>, <2 x i32>, <2 x i32> } [[VLD3_LANE]], { <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.int32x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.int32x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.int32x2x3_t, %struct.int32x2x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int32x2x3_t [[TMP16]]
 int32x2x3_t test_vld3_lane_s32(int32_t  *a, int32x2x3_t b) {
   return vld3_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X1X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X1X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x i64>] [[B_COERCE:%.*]], [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3lane.v1i64.p0i8(<1 x i64> [[TMP10]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3_LANE]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.int64x1x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.int64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT64X1X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x1x3_t @test_vld3_lane_s64(i64* %a, [3 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x1x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int64x1x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int64x1x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int64x1x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3lane.v1i64.p0i8(<1 x i64> [[TMP10]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3_LANE]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.int64x1x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.int64x1x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.int64x1x3_t, %struct.int64x1x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int64x1x3_t [[TMP16]]
 int64x1x3_t test_vld3_lane_s64(int64_t  *a, int64x1x3_t b) {
   return vld3_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld3_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X4X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X4X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x half>] [[B_COERCE:%.*]], [3 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x half> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x half> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x half>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x half>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x half>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3lane.v4f16.p0i8(<4 x half> [[TMP10]], <4 x half> [[TMP11]], <4 x half> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x half>, <4 x half>, <4 x half> }*
-// CHECK-NEXT:    store { <4 x half>, <4 x half>, <4 x half> } [[VLD3_LANE]], { <4 x half>, <4 x half>, <4 x half> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.float16x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.float16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT16X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float16x4x3_t @test_vld3_lane_f16(half* %a, [3 x <4 x half>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float16x4x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float16x4x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float16x4x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.float16x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x half> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x half> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x half> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x half>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x half>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x half>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3lane.v4f16.p0i8(<4 x half> [[TMP10]], <4 x half> [[TMP11]], <4 x half> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x half>, <4 x half>, <4 x half> }*
+// CHECK:   store { <4 x half>, <4 x half>, <4 x half> } [[VLD3_LANE]], { <4 x half>, <4 x half>, <4 x half> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.float16x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.float16x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.float16x4x3_t, %struct.float16x4x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float16x4x3_t [[TMP16]]
 float16x4x3_t test_vld3_lane_f16(float16_t  *a, float16x4x3_t b) {
   return vld3_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT32X2X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X2X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X2X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT32X2X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x float>] [[B_COERCE:%.*]], [3 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float32x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x float> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x float> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x float>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x float>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x float>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <2 x float>, <2 x float>, <2 x float> } @llvm.aarch64.neon.ld3lane.v2f32.p0i8(<2 x float> [[TMP10]], <2 x float> [[TMP11]], <2 x float> [[TMP12]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x float>, <2 x float>, <2 x float> }*
-// CHECK-NEXT:    store { <2 x float>, <2 x float>, <2 x float> } [[VLD3_LANE]], { <2 x float>, <2 x float>, <2 x float> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.float32x2x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.float32x2x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT32X2X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float32x2x3_t @test_vld3_lane_f32(float* %a, [3 x <2 x float>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float32x2x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float32x2x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float32x2x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.float32x2x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x float> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x float> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x float>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x float>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x float>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <2 x float>, <2 x float>, <2 x float> } @llvm.aarch64.neon.ld3lane.v2f32.p0i8(<2 x float> [[TMP10]], <2 x float> [[TMP11]], <2 x float> [[TMP12]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <2 x float>, <2 x float>, <2 x float> }*
+// CHECK:   store { <2 x float>, <2 x float>, <2 x float> } [[VLD3_LANE]], { <2 x float>, <2 x float>, <2 x float> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.float32x2x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.float32x2x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.float32x2x3_t, %struct.float32x2x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float32x2x3_t [[TMP16]]
 float32x2x3_t test_vld3_lane_f32(float32_t  *a, float32x2x3_t b) {
   return vld3_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld3_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x double>] [[B_COERCE:%.*]], [3 x <1 x double>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x double> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x double> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x double> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x double>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x double>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x double>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld3lane.v1f64.p0i8(<1 x double> [[TMP10]], <1 x double> [[TMP11]], <1 x double> [[TMP12]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x double>, <1 x double>, <1 x double> }*
-// CHECK-NEXT:    store { <1 x double>, <1 x double>, <1 x double> } [[VLD3_LANE]], { <1 x double>, <1 x double>, <1 x double> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.float64x1x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X1X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x1x3_t @test_vld3_lane_f64(double* %a, [3 x <1 x double>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x1x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float64x1x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float64x1x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP4:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x double> [[TMP4]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP6:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x double> [[TMP6]] to <8 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP8:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x double> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x double>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x double>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x double>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld3lane.v1f64.p0i8(<1 x double> [[TMP10]], <1 x double> [[TMP11]], <1 x double> [[TMP12]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x double>, <1 x double>, <1 x double> }*
+// CHECK:   store { <1 x double>, <1 x double>, <1 x double> } [[VLD3_LANE]], { <1 x double>, <1 x double>, <1 x double> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.float64x1x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.float64x1x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.float64x1x3_t, %struct.float64x1x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float64x1x3_t [[TMP16]]
 float64x1x3_t test_vld3_lane_f64(float64_t  *a, float64x1x3_t b) {
   return vld3_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld3_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY8X8X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X8X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X8X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY8X8X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i8>] [[B_COERCE:%.*]], [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly8x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld3lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8>, <8 x i8> } [[VLD3_LANE]], { <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP6]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast %struct.poly8x8x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.poly8x8x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 [[TMP8]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP9:%.*]] = load [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY8X8X3_T]] [[TMP9]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly8x8x3_t @test_vld3_lane_p8(i8* %a, [3 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly8x8x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly8x8x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x8x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly8x8x3_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
+// CHECK:   [[VLD3_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld3lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* %a)
+// CHECK:   [[TMP6:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8>, <8 x i8> } [[VLD3_LANE]], { <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP6]]
+// CHECK:   [[TMP7:%.*]] = bitcast %struct.poly8x8x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.poly8x8x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 [[TMP8]], i64 24, i1 false)
+// CHECK:   [[TMP9:%.*]] = load %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly8x8x3_t [[TMP9]]
 poly8x8x3_t test_vld3_lane_p8(poly8_t  *a, poly8x8x3_t b) {
   return vld3_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld3_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X4X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X4X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i16>] [[B_COERCE:%.*]], [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3lane.v4i16.p0i8(<4 x i16> [[TMP10]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16>, <4 x i16> } [[VLD3_LANE]], { <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.poly16x4x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.poly16x4x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY16X4X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly16x4x3_t @test_vld3_lane_p16(i16* %a, [3 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly16x4x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly16x4x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x4x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.poly16x4x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3lane.v4i16.p0i8(<4 x i16> [[TMP10]], <4 x i16> [[TMP11]], <4 x i16> [[TMP12]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16>, <4 x i16> } [[VLD3_LANE]], { <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.poly16x4x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.poly16x4x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly16x4x3_t [[TMP16]]
 poly16x4x3_t test_vld3_lane_p16(poly16_t  *a, poly16x4x3_t b) {
   return vld3_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld3_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X1X3_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X1X3_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x i64>] [[B_COERCE:%.*]], [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
-// CHECK-NEXT:    [[VLD3_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3lane.v1i64.p0i8(<1 x i64> [[TMP10]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3_LANE]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP13]], align 8
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast %struct.poly64x1x3_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP16:%.*]] = load [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY64X1X3_T]] [[TMP16]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x1x3_t @test_vld3_lane_p64(i64* %a, [3 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x3_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly64x1x3_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x3_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_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:   [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
+// CHECK:   [[VLD3_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld3lane.v1i64.p0i8(<1 x i64> [[TMP10]], <1 x i64> [[TMP11]], <1 x i64> [[TMP12]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP13:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64> } [[VLD3_LANE]], { <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP13]]
+// CHECK:   [[TMP14:%.*]] = bitcast %struct.poly64x1x3_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP15:%.*]] = bitcast %struct.poly64x1x3_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP14]], i8* align 8 [[TMP15]], i64 24, i1 false)
+// CHECK:   [[TMP16:%.*]] = load %struct.poly64x1x3_t, %struct.poly64x1x3_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly64x1x3_t [[TMP16]]
 poly64x1x3_t test_vld3_lane_p64(poly64_t  *a, poly64x1x3_t b) {
   return vld3_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT8X16X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X16X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X16X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT8X16X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <16 x i8>] [[B_COERCE:%.*]], [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x16x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x16x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint8x16x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], <16 x i8> [[TMP6]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[VLD4_LANE]], { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP7]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.uint8x16x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast %struct.uint8x16x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP8]], i8* align 16 [[TMP9]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP10:%.*]] = load [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT8X16X4_T]] [[TMP10]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint8x16x4_t @test_vld4q_lane_u8(i8* %a, [4 x <16 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint8x16x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint8x16x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x16x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.uint8x16x4_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = 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:   [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
+// CHECK:   [[VLD4_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], <16 x i8> [[TMP6]], i64 15, i8* %a)
+// CHECK:   [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[VLD4_LANE]], { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP7]]
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.uint8x16x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP9:%.*]] = bitcast %struct.uint8x16x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP8]], i8* align 16 [[TMP9]], i64 64, i1 false)
+// CHECK:   [[TMP10:%.*]] = load %struct.uint8x16x4_t, %struct.uint8x16x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint8x16x4_t [[TMP10]]
 uint8x16x4_t test_vld4q_lane_u8(uint8_t  *a, uint8x16x4_t b) {
   return vld4q_lane_u8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X8X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X8X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i16>] [[B_COERCE:%.*]], [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i16> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x i16>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld4lane.v8i16.p0i8(<8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], <8 x i16> [[TMP15]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } [[VLD4_LANE]], { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.uint16x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.uint16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT16X8X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint16x8x4_t @test_vld4q_lane_u16(i16* %a, [4 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint16x8x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint16x8x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x8x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint16x8x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] 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:   [[TMP10:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i16> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x i16>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld4lane.v8i16.p0i8(<8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], <8 x i16> [[TMP15]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } [[VLD4_LANE]], { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.uint16x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.uint16x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint16x8x4_t [[TMP19]]
 uint16x8x4_t test_vld4q_lane_u16(uint16_t  *a, uint16x8x4_t b) {
   return vld4q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT32X4X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X4X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X4X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT32X4X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i32>] [[B_COERCE:%.*]], [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint32x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x i32> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <4 x i32>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld4lane.v4i32.p0i8(<4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], <4 x i32> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }*
-// CHECK-NEXT:    store { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLD4_LANE]], { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.uint32x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.uint32x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT32X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint32x4x4_t @test_vld4q_lane_u32(i32* %a, [4 x <4 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint32x4x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint32x4x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x4x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint32x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x i32> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <4 x i32>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld4lane.v4i32.p0i8(<4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], <4 x i32> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }*
+// CHECK:   store { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLD4_LANE]], { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.uint32x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.uint32x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint32x4x4_t [[TMP19]]
 uint32x4x4_t test_vld4q_lane_u32(uint32_t  *a, uint32x4x4_t b) {
   return vld4q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X2X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X2X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i64>] [[B_COERCE:%.*]], [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x i64> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4lane.v2i64.p0i8(<2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], <2 x i64> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4_LANE]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.uint64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_UINT64X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x2x4_t @test_vld4q_lane_u64(i64* %a, [4 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x2x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.uint64x2x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] 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:   [[TMP10:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x i64> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4lane.v2i64.p0i8(<2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], <2 x i64> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4_LANE]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.uint64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.uint64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.uint64x2x4_t, %struct.uint64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.uint64x2x4_t [[TMP19]]
 uint64x2x4_t test_vld4q_lane_u64(uint64_t  *a, uint64x2x4_t b) {
   return vld4q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT8X16X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X16X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X16X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT8X16X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <16 x i8>] [[B_COERCE:%.*]], [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x16x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x16x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int8x16x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], <16 x i8> [[TMP6]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[VLD4_LANE]], { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP7]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.int8x16x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast %struct.int8x16x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP8]], i8* align 16 [[TMP9]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP10:%.*]] = load [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT8X16X4_T]] [[TMP10]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int8x16x4_t @test_vld4q_lane_s8(i8* %a, [4 x <16 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int8x16x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int8x16x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int8x16x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.int8x16x4_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = 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:   [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
+// CHECK:   [[VLD4_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], <16 x i8> [[TMP6]], i64 15, i8* %a)
+// CHECK:   [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[VLD4_LANE]], { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP7]]
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.int8x16x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP9:%.*]] = bitcast %struct.int8x16x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP8]], i8* align 16 [[TMP9]], i64 64, i1 false)
+// CHECK:   [[TMP10:%.*]] = load %struct.int8x16x4_t, %struct.int8x16x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int8x16x4_t [[TMP10]]
 int8x16x4_t test_vld4q_lane_s8(int8_t  *a, int8x16x4_t b) {
   return vld4q_lane_s8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X8X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X8X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i16>] [[B_COERCE:%.*]], [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i16> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x i16>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld4lane.v8i16.p0i8(<8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], <8 x i16> [[TMP15]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } [[VLD4_LANE]], { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.int16x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.int16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT16X8X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int16x8x4_t @test_vld4q_lane_s16(i16* %a, [4 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int16x8x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int16x8x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int16x8x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int16x8x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] 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:   [[TMP10:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i16> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x i16>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld4lane.v8i16.p0i8(<8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], <8 x i16> [[TMP15]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } [[VLD4_LANE]], { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.int16x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.int16x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.int16x8x4_t, %struct.int16x8x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int16x8x4_t [[TMP19]]
 int16x8x4_t test_vld4q_lane_s16(int16_t  *a, int16x8x4_t b) {
   return vld4q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT32X4X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X4X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X4X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT32X4X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i32>] [[B_COERCE:%.*]], [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int32x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x i32> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <4 x i32>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld4lane.v4i32.p0i8(<4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], <4 x i32> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }*
-// CHECK-NEXT:    store { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLD4_LANE]], { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.int32x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.int32x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT32X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int32x4x4_t @test_vld4q_lane_s32(i32* %a, [4 x <4 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int32x4x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int32x4x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int32x4x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int32x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i32> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x i32> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x i32>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x i32>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x i32>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <4 x i32>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld4lane.v4i32.p0i8(<4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], <4 x i32> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }*
+// CHECK:   store { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> } [[VLD4_LANE]], { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.int32x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.int32x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.int32x4x4_t, %struct.int32x4x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int32x4x4_t [[TMP19]]
 int32x4x4_t test_vld4q_lane_s32(int32_t  *a, int32x4x4_t b) {
   return vld4q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X2X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X2X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i64>] [[B_COERCE:%.*]], [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x i64> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4lane.v2i64.p0i8(<2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], <2 x i64> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4_LANE]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.int64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_INT64X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x2x4_t @test_vld4q_lane_s64(i64* %a, [4 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x2x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.int64x2x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.int64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.int64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] 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:   [[TMP10:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x i64> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4lane.v2i64.p0i8(<2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], <2 x i64> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4_LANE]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.int64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.int64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.int64x2x4_t, %struct.int64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.int64x2x4_t [[TMP19]]
 int64x2x4_t test_vld4q_lane_s64(int64_t  *a, int64x2x4_t b) {
   return vld4q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X8X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X8X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x half>] [[B_COERCE:%.*]], [4 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x half> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x half> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x half> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x half> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x half>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x half>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x half>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x half>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4lane.v8f16.p0i8(<8 x half> [[TMP12]], <8 x half> [[TMP13]], <8 x half> [[TMP14]], <8 x half> [[TMP15]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x half>, <8 x half>, <8 x half>, <8 x half> }*
-// CHECK-NEXT:    store { <8 x half>, <8 x half>, <8 x half>, <8 x half> } [[VLD4_LANE]], { <8 x half>, <8 x half>, <8 x half>, <8 x half> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.float16x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.float16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT16X8X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float16x8x4_t @test_vld4q_lane_f16(half* %a, [4 x <8 x half>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float16x8x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float16x8x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float16x8x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.float16x8x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x half> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x half> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x half> [[TMP8]] 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:   [[TMP10:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x half> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x half>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x half>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x half>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x half>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4lane.v8f16.p0i8(<8 x half> [[TMP12]], <8 x half> [[TMP13]], <8 x half> [[TMP14]], <8 x half> [[TMP15]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x half>, <8 x half>, <8 x half>, <8 x half> }*
+// CHECK:   store { <8 x half>, <8 x half>, <8 x half>, <8 x half> } [[VLD4_LANE]], { <8 x half>, <8 x half>, <8 x half>, <8 x half> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.float16x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.float16x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.float16x8x4_t, %struct.float16x8x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float16x8x4_t [[TMP19]]
 float16x8x4_t test_vld4q_lane_f16(float16_t  *a, float16x8x4_t b) {
   return vld4q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT32X4X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X4X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X4X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT32X4X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x float>] [[B_COERCE:%.*]], [4 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float32x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x float> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x float> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x float> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x float> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x float>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x float>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x float>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <4 x float>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x float>, <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld4lane.v4f32.p0i8(<4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]], <4 x float> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x float>, <4 x float>, <4 x float>, <4 x float> }*
-// CHECK-NEXT:    store { <4 x float>, <4 x float>, <4 x float>, <4 x float> } [[VLD4_LANE]], { <4 x float>, <4 x float>, <4 x float>, <4 x float> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.float32x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.float32x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT32X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float32x4x4_t @test_vld4q_lane_f32(float* %a, [4 x <4 x float>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float32x4x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float32x4x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float32x4x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.float32x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x float> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x float> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x float> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x float> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <4 x float>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <4 x float>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <4 x float>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <4 x float>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x float>, <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld4lane.v4f32.p0i8(<4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]], <4 x float> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x float>, <4 x float>, <4 x float>, <4 x float> }*
+// CHECK:   store { <4 x float>, <4 x float>, <4 x float>, <4 x float> } [[VLD4_LANE]], { <4 x float>, <4 x float>, <4 x float>, <4 x float> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.float32x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.float32x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.float32x4x4_t, %struct.float32x4x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float32x4x4_t [[TMP19]]
 float32x4x4_t test_vld4q_lane_f32(float32_t  *a, float32x4x4_t b) {
   return vld4q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x double>] [[B_COERCE:%.*]], [4 x <2 x double>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x double> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x double> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x double> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x double> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x double>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x double>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x double>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x double>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x double>, <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld4lane.v2f64.p0i8(<2 x double> [[TMP12]], <2 x double> [[TMP13]], <2 x double> [[TMP14]], <2 x double> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x double>, <2 x double>, <2 x double>, <2 x double> }*
-// CHECK-NEXT:    store { <2 x double>, <2 x double>, <2 x double>, <2 x double> } [[VLD4_LANE]], { <2 x double>, <2 x double>, <2 x double>, <2 x double> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.float64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x2x4_t @test_vld4q_lane_f64(double* %a, [4 x <2 x double>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x2x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.float64x2x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.float64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x double> [[TMP4]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP6:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x double> [[TMP6]] to <16 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP8:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x double> [[TMP8]] to <16 x i8>
+// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL5]], i64 0, i64 3
+// CHECK:   [[TMP10:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x double> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x double>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x double>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x double>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x double>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x double>, <2 x double>, <2 x double>, <2 x double> } @llvm.aarch64.neon.ld4lane.v2f64.p0i8(<2 x double> [[TMP12]], <2 x double> [[TMP13]], <2 x double> [[TMP14]], <2 x double> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x double>, <2 x double>, <2 x double>, <2 x double> }*
+// CHECK:   store { <2 x double>, <2 x double>, <2 x double>, <2 x double> } [[VLD4_LANE]], { <2 x double>, <2 x double>, <2 x double>, <2 x double> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.float64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.float64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.float64x2x4_t, %struct.float64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.float64x2x4_t [[TMP19]]
 float64x2x4_t test_vld4q_lane_f64(float64_t  *a, float64x2x4_t b) {
   return vld4q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY8X16X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X16X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X16X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY8X16X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <16 x i8>] [[B_COERCE:%.*]], [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x16x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x16x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly8x16x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], <16 x i8> [[TMP6]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }*
-// CHECK-NEXT:    store { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[VLD4_LANE]], { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP7]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.poly8x16x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast %struct.poly8x16x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP8]], i8* align 16 [[TMP9]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP10:%.*]] = load [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY8X16X4_T]] [[TMP10]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly8x16x4_t @test_vld4q_lane_p8(i8* %a, [4 x <16 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly8x16x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly8x16x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x16x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly8x16x4_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = 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:   [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
+// CHECK:   [[VLD4_LANE:%.*]] = call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4lane.v16i8.p0i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], <16 x i8> [[TMP6]], i64 15, i8* %a)
+// CHECK:   [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }*
+// CHECK:   store { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[VLD4_LANE]], { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }* [[TMP7]]
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.poly8x16x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP9:%.*]] = bitcast %struct.poly8x16x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP8]], i8* align 16 [[TMP9]], i64 64, i1 false)
+// CHECK:   [[TMP10:%.*]] = load %struct.poly8x16x4_t, %struct.poly8x16x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly8x16x4_t [[TMP10]]
 poly8x16x4_t test_vld4q_lane_p8(poly8_t  *a, poly8x16x4_t b) {
   return vld4q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X8X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X8X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i16>] [[B_COERCE:%.*]], [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i16> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x i16>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld4lane.v8i16.p0i8(<8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], <8 x i16> [[TMP15]], i64 7, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }*
-// CHECK-NEXT:    store { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } [[VLD4_LANE]], { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.poly16x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.poly16x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY16X8X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly16x8x4_t @test_vld4q_lane_p16(i16* %a, [4 x <8 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly16x8x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly16x8x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x8x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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 %struct.poly16x8x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <8 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i16> [[TMP8]] 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:   [[TMP10:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i16> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <8 x i16>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <8 x i16>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <8 x i16>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <8 x i16>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } @llvm.aarch64.neon.ld4lane.v8i16.p0i8(<8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], <8 x i16> [[TMP15]], i64 7, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }*
+// CHECK:   store { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> } [[VLD4_LANE]], { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.poly16x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.poly16x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly16x8x4_t [[TMP19]]
 poly16x8x4_t test_vld4q_lane_p16(poly16_t  *a, poly16x8x4_t b) {
   return vld4q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X2X4_T]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X2X4_T]], align 16
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i64>] [[B_COERCE:%.*]], [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x i64> [[TMP10]] to <16 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4lane.v2i64.p0i8(<2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], <2 x i64> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
-// CHECK-NEXT:    store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4_LANE]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP16]], align 16
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.poly64x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[RETVAL]], align 16
-// CHECK-NEXT:    ret [[STRUCT_POLY64X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x2x4_t @test_vld4q_lane_p64(i64* %a, [4 x <2 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x2x4_t, align 16
+// CHECK:   [[B:%.*]] = alloca %struct.poly64x2x4_t, align 16
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly64x2x4_t, align 16
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_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:   [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i64> [[TMP4]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_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:   [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i64> [[TMP6]] to <16 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_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:   [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i64> [[TMP8]] to <16 x i8>
+// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_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:   [[TMP10:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x i64> [[TMP10]] to <16 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP5]] to <2 x i64>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP7]] to <2 x i64>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP9]] to <2 x i64>
+// CHECK:   [[TMP15:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4lane.v2i64.p0i8(<2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], <2 x i64> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }*
+// CHECK:   store { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } [[VLD4_LANE]], { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.poly64x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.poly64x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP17]], i8* align 16 [[TMP18]], i64 64, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.poly64x2x4_t, %struct.poly64x2x4_t* [[RETVAL]], align 16
+// CHECK:   ret %struct.poly64x2x4_t [[TMP19]]
 poly64x2x4_t test_vld4q_lane_p64(poly64_t  *a, poly64x2x4_t b) {
   return vld4q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT8X8X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X8X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X8X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT8X8X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i8>] [[B_COERCE:%.*]], [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint8x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld4lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], <8 x i8> [[TMP6]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } [[VLD4_LANE]], { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP7]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.uint8x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast %struct.uint8x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 [[TMP9]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP10:%.*]] = load [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT8X8X4_T]] [[TMP10]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint8x8x4_t @test_vld4_lane_u8(i8* %a, [4 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint8x8x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint8x8x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint8x8x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.uint8x8x4_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = 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:   [[TMP6:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld4lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], <8 x i8> [[TMP6]], i64 7, i8* %a)
+// CHECK:   [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } [[VLD4_LANE]], { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP7]]
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.uint8x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP9:%.*]] = bitcast %struct.uint8x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 [[TMP9]], i64 32, i1 false)
+// CHECK:   [[TMP10:%.*]] = load %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint8x8x4_t [[TMP10]]
 uint8x8x4_t test_vld4_lane_u8(uint8_t  *a, uint8x8x4_t b) {
   return vld4_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X4X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X4X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i16>] [[B_COERCE:%.*]], [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x i16> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x i16>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4lane.v4i16.p0i8(<4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], <4 x i16> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } [[VLD4_LANE]], { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.uint16x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.uint16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT16X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint16x4x4_t @test_vld4_lane_u16(i16* %a, [4 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint16x4x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint16x4x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint16x4x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint16x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x i16> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x i16>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4lane.v4i16.p0i8(<4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], <4 x i16> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } [[VLD4_LANE]], { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.uint16x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.uint16x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint16x4x4_t [[TMP19]]
 uint16x4x4_t test_vld4_lane_u16(uint16_t  *a, uint16x4x4_t b) {
   return vld4_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT32X2X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X2X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X2X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT32X2X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i32>] [[B_COERCE:%.*]], [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint32x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x i32> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <2 x i32>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld4lane.v2i32.p0i8(<2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], <2 x i32> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }*
-// CHECK-NEXT:    store { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } [[VLD4_LANE]], { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.uint32x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.uint32x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT32X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint32x2x4_t @test_vld4_lane_u32(i32* %a, [4 x <2 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint32x2x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint32x2x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint32x2x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint32x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] 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:   [[TMP10:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x i32> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <2 x i32>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld4lane.v2i32.p0i8(<2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], <2 x i32> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }*
+// CHECK:   store { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } [[VLD4_LANE]], { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.uint32x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.uint32x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint32x2x4_t [[TMP19]]
 uint32x2x4_t test_vld4_lane_u32(uint32_t  *a, uint32x2x4_t b) {
   return vld4_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_UINT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X1X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X1X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_UINT64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x i64>] [[B_COERCE:%.*]], [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.uint64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <1 x i64> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x i64>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4lane.v1i64.p0i8(<1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], <1 x i64> [[TMP15]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4_LANE]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.uint64x1x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.uint64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_UINT64X1X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.uint64x1x4_t @test_vld4_lane_u64(i64* %a, [4 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.uint64x1x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.uint64x1x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.uint64x1x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.uint64x1x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] 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:   [[TMP10:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <1 x i64> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x i64>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4lane.v1i64.p0i8(<1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], <1 x i64> [[TMP15]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4_LANE]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.uint64x1x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.uint64x1x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.uint64x1x4_t, %struct.uint64x1x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.uint64x1x4_t [[TMP19]]
 uint64x1x4_t test_vld4_lane_u64(uint64_t  *a, uint64x1x4_t b) {
   return vld4_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld4_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT8X8X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X8X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X8X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT8X8X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i8>] [[B_COERCE:%.*]], [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int8x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld4lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], <8 x i8> [[TMP6]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } [[VLD4_LANE]], { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP7]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.int8x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast %struct.int8x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 [[TMP9]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP10:%.*]] = load [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT8X8X4_T]] [[TMP10]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int8x8x4_t @test_vld4_lane_s8(i8* %a, [4 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int8x8x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int8x8x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int8x8x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.int8x8x4_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = 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:   [[TMP6:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld4lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], <8 x i8> [[TMP6]], i64 7, i8* %a)
+// CHECK:   [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } [[VLD4_LANE]], { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP7]]
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.int8x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP9:%.*]] = bitcast %struct.int8x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 [[TMP9]], i64 32, i1 false)
+// CHECK:   [[TMP10:%.*]] = load %struct.int8x8x4_t, %struct.int8x8x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int8x8x4_t [[TMP10]]
 int8x8x4_t test_vld4_lane_s8(int8_t  *a, int8x8x4_t b) {
   return vld4_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X4X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X4X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i16>] [[B_COERCE:%.*]], [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x i16> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x i16>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4lane.v4i16.p0i8(<4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], <4 x i16> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } [[VLD4_LANE]], { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.int16x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.int16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT16X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int16x4x4_t @test_vld4_lane_s16(i16* %a, [4 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int16x4x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int16x4x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int16x4x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int16x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x i16> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x i16>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4lane.v4i16.p0i8(<4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], <4 x i16> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } [[VLD4_LANE]], { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.int16x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.int16x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.int16x4x4_t, %struct.int16x4x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int16x4x4_t [[TMP19]]
 int16x4x4_t test_vld4_lane_s16(int16_t  *a, int16x4x4_t b) {
   return vld4_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT32X2X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X2X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X2X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT32X2X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i32>] [[B_COERCE:%.*]], [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int32x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x i32> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <2 x i32>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld4lane.v2i32.p0i8(<2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], <2 x i32> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }*
-// CHECK-NEXT:    store { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } [[VLD4_LANE]], { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.int32x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.int32x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT32X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int32x2x4_t @test_vld4_lane_s32(i32* %a, [4 x <2 x i32>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int32x2x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int32x2x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int32x2x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int32x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x i32> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x i32> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x i32> [[TMP8]] 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:   [[TMP10:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x i32> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x i32>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x i32>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x i32>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <2 x i32>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld4lane.v2i32.p0i8(<2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], <2 x i32> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }*
+// CHECK:   store { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> } [[VLD4_LANE]], { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.int32x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.int32x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.int32x2x4_t, %struct.int32x2x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int32x2x4_t [[TMP19]]
 int32x2x4_t test_vld4_lane_s32(int32_t  *a, int32x2x4_t b) {
   return vld4_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_INT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X1X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X1X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_INT64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x i64>] [[B_COERCE:%.*]], [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.int64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <1 x i64> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x i64>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4lane.v1i64.p0i8(<1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], <1 x i64> [[TMP15]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4_LANE]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.int64x1x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.int64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_INT64X1X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.int64x1x4_t @test_vld4_lane_s64(i64* %a, [4 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.int64x1x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.int64x1x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.int64x1x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.int64x1x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] 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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] 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:   [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] 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:   [[TMP10:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <1 x i64> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x i64>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4lane.v1i64.p0i8(<1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], <1 x i64> [[TMP15]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4_LANE]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.int64x1x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.int64x1x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.int64x1x4_t, %struct.int64x1x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.int64x1x4_t [[TMP19]]
 int64x1x4_t test_vld4_lane_s64(int64_t  *a, int64x1x4_t b) {
   return vld4_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld4_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X4X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X4X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x half>] [[B_COERCE:%.*]], [4 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x half> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x half> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x half> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x half> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x half>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x half>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x half>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x half>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4lane.v4f16.p0i8(<4 x half> [[TMP12]], <4 x half> [[TMP13]], <4 x half> [[TMP14]], <4 x half> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x half>, <4 x half>, <4 x half>, <4 x half> }*
-// CHECK-NEXT:    store { <4 x half>, <4 x half>, <4 x half>, <4 x half> } [[VLD4_LANE]], { <4 x half>, <4 x half>, <4 x half>, <4 x half> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.float16x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.float16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT16X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float16x4x4_t @test_vld4_lane_f16(half* %a, [4 x <4 x half>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float16x4x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float16x4x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float16x4x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.float16x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x half> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x half> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x half> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x half> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x half>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x half>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x half>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x half>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4lane.v4f16.p0i8(<4 x half> [[TMP12]], <4 x half> [[TMP13]], <4 x half> [[TMP14]], <4 x half> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x half>, <4 x half>, <4 x half>, <4 x half> }*
+// CHECK:   store { <4 x half>, <4 x half>, <4 x half>, <4 x half> } [[VLD4_LANE]], { <4 x half>, <4 x half>, <4 x half>, <4 x half> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.float16x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.float16x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.float16x4x4_t, %struct.float16x4x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float16x4x4_t [[TMP19]]
 float16x4x4_t test_vld4_lane_f16(float16_t  *a, float16x4x4_t b) {
   return vld4_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT32X2X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X2X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X2X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT32X2X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x float>] [[B_COERCE:%.*]], [4 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float32x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <2 x float> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <2 x float> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <2 x float> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x float>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x float>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x float>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <2 x float>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <2 x float>, <2 x float>, <2 x float>, <2 x float> } @llvm.aarch64.neon.ld4lane.v2f32.p0i8(<2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]], <2 x float> [[TMP15]], i64 1, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x float>, <2 x float>, <2 x float>, <2 x float> }*
-// CHECK-NEXT:    store { <2 x float>, <2 x float>, <2 x float>, <2 x float> } [[VLD4_LANE]], { <2 x float>, <2 x float>, <2 x float>, <2 x float> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.float32x2x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.float32x2x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT32X2X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float32x2x4_t @test_vld4_lane_f32(float* %a, [4 x <2 x float>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float32x2x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float32x2x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float32x2x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.float32x2x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <2 x float> [[TMP4]] 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:   [[TMP6:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <2 x float> [[TMP6]] 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:   [[TMP8:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <2 x float> [[TMP8]] 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:   [[TMP10:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <2 x float> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <2 x float>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <2 x float>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <2 x float>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <2 x float>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <2 x float>, <2 x float>, <2 x float>, <2 x float> } @llvm.aarch64.neon.ld4lane.v2f32.p0i8(<2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]], <2 x float> [[TMP15]], i64 1, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <2 x float>, <2 x float>, <2 x float>, <2 x float> }*
+// CHECK:   store { <2 x float>, <2 x float>, <2 x float>, <2 x float> } [[VLD4_LANE]], { <2 x float>, <2 x float>, <2 x float>, <2 x float> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.float32x2x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.float32x2x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.float32x2x4_t, %struct.float32x2x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float32x2x4_t [[TMP19]]
 float32x2x4_t test_vld4_lane_f32(float32_t  *a, float32x2x4_t b) {
   return vld4_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vld4_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x double>] [[B_COERCE:%.*]], [4 x <1 x double>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x double> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x double> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x double> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <1 x double> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x double>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x double>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x double>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x double>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <1 x double>, <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld4lane.v1f64.p0i8(<1 x double> [[TMP12]], <1 x double> [[TMP13]], <1 x double> [[TMP14]], <1 x double> [[TMP15]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x double>, <1 x double>, <1 x double>, <1 x double> }*
-// CHECK-NEXT:    store { <1 x double>, <1 x double>, <1 x double>, <1 x double> } [[VLD4_LANE]], { <1 x double>, <1 x double>, <1 x double>, <1 x double> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.float64x1x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_FLOAT64X1X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.float64x1x4_t @test_vld4_lane_f64(double* %a, [4 x <1 x double>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.float64x1x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.float64x1x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.float64x1x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP4:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x double> [[TMP4]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP6:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x double> [[TMP6]] to <8 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP8:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x double> [[TMP8]] to <8 x i8>
+// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL5]], i64 0, i64 3
+// CHECK:   [[TMP10:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <1 x double> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x double>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x double>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x double>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x double>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <1 x double>, <1 x double>, <1 x double>, <1 x double> } @llvm.aarch64.neon.ld4lane.v1f64.p0i8(<1 x double> [[TMP12]], <1 x double> [[TMP13]], <1 x double> [[TMP14]], <1 x double> [[TMP15]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x double>, <1 x double>, <1 x double>, <1 x double> }*
+// CHECK:   store { <1 x double>, <1 x double>, <1 x double>, <1 x double> } [[VLD4_LANE]], { <1 x double>, <1 x double>, <1 x double>, <1 x double> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.float64x1x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.float64x1x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.float64x1x4_t, %struct.float64x1x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.float64x1x4_t [[TMP19]]
 float64x1x4_t test_vld4_lane_f64(float64_t  *a, float64x1x4_t b) {
   return vld4_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vld4_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY8X8X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X8X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X8X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY8X8X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i8>] [[B_COERCE:%.*]], [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly8x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP6:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld4lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], <8 x i8> [[TMP6]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }*
-// CHECK-NEXT:    store { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } [[VLD4_LANE]], { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP7]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast %struct.poly8x8x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast %struct.poly8x8x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 [[TMP9]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP10:%.*]] = load [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY8X8X4_T]] [[TMP10]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly8x8x4_t @test_vld4_lane_p8(i8* %a, [4 x <8 x i8>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly8x8x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly8x8x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly8x8x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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:   [[TMP2:%.*]] = bitcast %struct.poly8x8x4_t* [[__RET]] to i8*
+// 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:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = 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:   [[TMP5:%.*]] = 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:   [[TMP6:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
+// CHECK:   [[VLD4_LANE:%.*]] = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.aarch64.neon.ld4lane.v8i8.p0i8(<8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], <8 x i8> [[TMP6]], i64 7, i8* %a)
+// CHECK:   [[TMP7:%.*]] = bitcast i8* [[TMP2]] to { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }*
+// CHECK:   store { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } [[VLD4_LANE]], { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }* [[TMP7]]
+// CHECK:   [[TMP8:%.*]] = bitcast %struct.poly8x8x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP9:%.*]] = bitcast %struct.poly8x8x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 [[TMP9]], i64 32, i1 false)
+// CHECK:   [[TMP10:%.*]] = load %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly8x8x4_t [[TMP10]]
 poly8x8x4_t test_vld4_lane_p8(poly8_t  *a, poly8x8x4_t b) {
   return vld4_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vld4_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X4X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X4X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i16>] [[B_COERCE:%.*]], [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <4 x i16> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x i16>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4lane.v4i16.p0i8(<4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], <4 x i16> [[TMP15]], i64 3, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }*
-// CHECK-NEXT:    store { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } [[VLD4_LANE]], { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.poly16x4x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.poly16x4x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY16X4X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly16x4x4_t @test_vld4_lane_p16(i16* %a, [4 x <4 x i16>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly16x4x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly16x4x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly16x4x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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 %struct.poly16x4x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = 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:   [[TMP4:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <4 x i16> [[TMP4]] 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:   [[TMP6:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <4 x i16> [[TMP6]] 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:   [[TMP8:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <4 x i16> [[TMP8]] 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:   [[TMP10:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <4 x i16> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <4 x i16>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <4 x i16>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <4 x i16>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <4 x i16>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4lane.v4i16.p0i8(<4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], <4 x i16> [[TMP15]], i64 3, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }*
+// CHECK:   store { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } [[VLD4_LANE]], { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.poly16x4x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.poly16x4x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly16x4x4_t [[TMP19]]
 poly16x4x4_t test_vld4_lane_p16(poly16_t  *a, poly16x4x4_t b) {
   return vld4_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vld4_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_POLY64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X1X4_T]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X1X4_T]], align 8
-// CHECK-NEXT:    [[__RET:%.*]] = alloca [[STRUCT_POLY64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x i64>] [[B_COERCE:%.*]], [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP10:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <1 x i64> [[TMP10]] to <8 x i8>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
-// CHECK-NEXT:    [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x i64>
-// CHECK-NEXT:    [[VLD4_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4lane.v1i64.p0i8(<1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], <1 x i64> [[TMP15]], i64 0, i8* [[TMP3]])
-// CHECK-NEXT:    [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
-// CHECK-NEXT:    store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4_LANE]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP16]], align 8
-// CHECK-NEXT:    [[TMP17:%.*]] = bitcast %struct.poly64x1x4_t* [[RETVAL]] to i8*
-// CHECK-NEXT:    [[TMP18:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP19:%.*]] = load [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[RETVAL]], align 8
-// CHECK-NEXT:    ret [[STRUCT_POLY64X1X4_T]] [[TMP19]]
-//
+// CHECK-LABEL: define{{.*}} %struct.poly64x1x4_t @test_vld4_lane_p64(i64* %a, [4 x <1 x i64>] %b.coerce) #2 {
+// CHECK:   [[RETVAL:%.*]] = alloca %struct.poly64x1x4_t, align 8
+// CHECK:   [[B:%.*]] = alloca %struct.poly64x1x4_t, align 8
+// CHECK:   [[__S1:%.*]] = alloca %struct.poly64x1x4_t, align 8
+// CHECK:   [[__RET:%.*]] = 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)
+// CHECK:   [[TMP2:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
+// CHECK:   [[TMP3:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_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:   [[TMP4:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP5:%.*]] = bitcast <1 x i64> [[TMP4]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_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:   [[TMP6:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP7:%.*]] = bitcast <1 x i64> [[TMP6]] to <8 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_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:   [[TMP8:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP9:%.*]] = bitcast <1 x i64> [[TMP8]] to <8 x i8>
+// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_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:   [[TMP10:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP11:%.*]] = bitcast <1 x i64> [[TMP10]] to <8 x i8>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP5]] to <1 x i64>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP7]] to <1 x i64>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP9]] to <1 x i64>
+// CHECK:   [[TMP15:%.*]] = bitcast <8 x i8> [[TMP11]] to <1 x i64>
+// CHECK:   [[VLD4_LANE:%.*]] = call { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } @llvm.aarch64.neon.ld4lane.v1i64.p0i8(<1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], <1 x i64> [[TMP15]], i64 0, i8* [[TMP3]])
+// CHECK:   [[TMP16:%.*]] = bitcast i8* [[TMP2]] to { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }*
+// CHECK:   store { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> } [[VLD4_LANE]], { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }* [[TMP16]]
+// CHECK:   [[TMP17:%.*]] = bitcast %struct.poly64x1x4_t* [[RETVAL]] to i8*
+// CHECK:   [[TMP18:%.*]] = bitcast %struct.poly64x1x4_t* [[__RET]] to i8*
+// CHECK:   call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP17]], i8* align 8 [[TMP18]], i64 32, i1 false)
+// CHECK:   [[TMP19:%.*]] = load %struct.poly64x1x4_t, %struct.poly64x1x4_t* [[RETVAL]], align 8
+// CHECK:   ret %struct.poly64x1x4_t [[TMP19]]
 poly64x1x4_t test_vld4_lane_p64(poly64_t  *a, poly64x1x4_t b) {
   return vld4_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = extractelement <16 x i8> [[B:%.*]], i32 15
-// CHECK-NEXT:    store i8 [[TMP0]], i8* [[A:%.*]], align 1
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_u8(i8* %a, <16 x i8> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = extractelement <16 x i8> %b, i32 15
+// CHECK:   store i8 [[TMP0]], i8* %a
+// CHECK:   ret void
 void test_vst1q_lane_u8(uint8_t  *a, uint8x16_t b) {
   vst1q_lane_u8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    store i16 [[TMP3]], i16* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_u16(i16* %a, <8 x i16> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// CHECK:   [[TMP3:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   store i16 [[TMP3]], i16* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_u16(uint16_t  *a, uint16x8_t b) {
   vst1q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_u32(i32* %a, <4 x i32> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i32> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   store i32 [[TMP3]], i32* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_u32(uint32_t  *a, uint32x4_t b) {
   vst1q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    store i64 [[TMP3]], i64* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_u64(i64* %a, <2 x i64> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   store i64 [[TMP3]], i64* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_u64(uint64_t  *a, uint64x2_t b) {
   vst1q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = extractelement <16 x i8> [[B:%.*]], i32 15
-// CHECK-NEXT:    store i8 [[TMP0]], i8* [[A:%.*]], align 1
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_s8(i8* %a, <16 x i8> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = extractelement <16 x i8> %b, i32 15
+// CHECK:   store i8 [[TMP0]], i8* %a
+// CHECK:   ret void
 void test_vst1q_lane_s8(int8_t  *a, int8x16_t b) {
   vst1q_lane_s8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    store i16 [[TMP3]], i16* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_s16(i16* %a, <8 x i16> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// CHECK:   [[TMP3:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   store i16 [[TMP3]], i16* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_s16(int16_t  *a, int16x8_t b) {
   vst1q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_s32(i32* %a, <4 x i32> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i32> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   store i32 [[TMP3]], i32* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_s32(int32_t  *a, int32x4_t b) {
   vst1q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    store i64 [[TMP3]], i64* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_s64(i64* %a, <2 x i64> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   store i64 [[TMP3]], i64* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_s64(int64_t  *a, int64x2_t b) {
   vst1q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x half> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x half>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x half> [[TMP2]], i32 7
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to half*
-// CHECK-NEXT:    store half [[TMP3]], half* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_f16(half* %a, <8 x half> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast half* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x half> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x half>
+// CHECK:   [[TMP3:%.*]] = extractelement <8 x half> [[TMP2]], i32 7
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to half*
+// CHECK:   store half [[TMP3]], half* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_f16(float16_t  *a, float16x8_t b) {
   vst1q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x float> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to float*
-// CHECK-NEXT:    store float [[TMP3]], float* [[TMP4]], align 4
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_f32(float* %a, <4 x float> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast float* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x float> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x float>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x float> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to float*
+// CHECK:   store float [[TMP3]], float* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_f32(float32_t  *a, float32x4_t b) {
   vst1q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x double> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x double> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to double*
-// CHECK-NEXT:    store double [[TMP3]], double* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_f64(double* %a, <2 x double> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x double> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x double>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x double> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to double*
+// CHECK:   store double [[TMP3]], double* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_f64(float64_t  *a, float64x2_t b) {
   vst1q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = extractelement <16 x i8> [[B:%.*]], i32 15
-// CHECK-NEXT:    store i8 [[TMP0]], i8* [[A:%.*]], align 1
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_p8(i8* %a, <16 x i8> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = extractelement <16 x i8> %b, i32 15
+// CHECK:   store i8 [[TMP0]], i8* %a
+// CHECK:   ret void
 void test_vst1q_lane_p8(poly8_t  *a, poly8x16_t b) {
   vst1q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    store i16 [[TMP3]], i16* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_p16(i16* %a, <8 x i16> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <8 x i16> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// CHECK:   [[TMP3:%.*]] = extractelement <8 x i16> [[TMP2]], i32 7
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   store i16 [[TMP3]], i16* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_p16(poly16_t  *a, poly16x8_t b) {
   vst1q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    store i64 [[TMP3]], i64* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1q_lane_p64(i64* %a, <2 x i64> %b) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i64> %b to <16 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x i64> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   store i64 [[TMP3]], i64* [[TMP4]]
+// CHECK:   ret void
 void test_vst1q_lane_p64(poly64_t  *a, poly64x2_t b) {
   vst1q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = extractelement <8 x i8> [[B:%.*]], i32 7
-// CHECK-NEXT:    store i8 [[TMP0]], i8* [[A:%.*]], align 1
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_u8(i8* %a, <8 x i8> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = extractelement <8 x i8> %b, i32 7
+// CHECK:   store i8 [[TMP0]], i8* %a
+// CHECK:   ret void
 void test_vst1_lane_u8(uint8_t  *a, uint8x8_t b) {
   vst1_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    store i16 [[TMP3]], i16* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_u16(i16* %a, <4 x i16> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   store i16 [[TMP3]], i16* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_u16(uint16_t  *a, uint16x4_t b) {
   vst1_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_u32(i32* %a, <2 x i32> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i32> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   store i32 [[TMP3]], i32* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_u32(uint32_t  *a, uint32x2_t b) {
   vst1_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <1 x i64> [[TMP2]], i32 0
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    store i64 [[TMP3]], i64* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_u64(i64* %a, <1 x i64> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// CHECK:   [[TMP3:%.*]] = extractelement <1 x i64> [[TMP2]], i32 0
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   store i64 [[TMP3]], i64* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_u64(uint64_t  *a, uint64x1_t b) {
   vst1_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst1_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = extractelement <8 x i8> [[B:%.*]], i32 7
-// CHECK-NEXT:    store i8 [[TMP0]], i8* [[A:%.*]], align 1
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_s8(i8* %a, <8 x i8> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = extractelement <8 x i8> %b, i32 7
+// CHECK:   store i8 [[TMP0]], i8* %a
+// CHECK:   ret void
 void test_vst1_lane_s8(int8_t  *a, int8x8_t b) {
   vst1_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    store i16 [[TMP3]], i16* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_s16(i16* %a, <4 x i16> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   store i16 [[TMP3]], i16* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_s16(int16_t  *a, int16x4_t b) {
   vst1_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
-// CHECK-NEXT:    store i32 [[TMP3]], i32* [[TMP4]], align 4
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_s32(i32* %a, <2 x i32> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i32* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x i32> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i32*
+// CHECK:   store i32 [[TMP3]], i32* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_s32(int32_t  *a, int32x2_t b) {
   vst1_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <1 x i64> [[TMP2]], i32 0
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    store i64 [[TMP3]], i64* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_s64(i64* %a, <1 x i64> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// CHECK:   [[TMP3:%.*]] = extractelement <1 x i64> [[TMP2]], i32 0
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   store i64 [[TMP3]], i64* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_s64(int64_t  *a, int64x1_t b) {
   vst1_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst1_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x half> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x half>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x half> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to half*
-// CHECK-NEXT:    store half [[TMP3]], half* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_f16(half* %a, <4 x half> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast half* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x half> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x half>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x half> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to half*
+// CHECK:   store half [[TMP3]], half* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_f16(float16_t  *a, float16x4_t b) {
   vst1_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x float> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to float*
-// CHECK-NEXT:    store float [[TMP3]], float* [[TMP4]], align 4
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_f32(float* %a, <2 x float> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast float* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <2 x float> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x float>
+// CHECK:   [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to float*
+// CHECK:   store float [[TMP3]], float* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_f32(float32_t  *a, float32x2_t b) {
   vst1_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst1_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <1 x double> [[TMP2]], i32 0
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to double*
-// CHECK-NEXT:    store double [[TMP3]], double* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_f64(double* %a, <1 x double> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x double> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
+// CHECK:   [[TMP3:%.*]] = extractelement <1 x double> [[TMP2]], i32 0
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to double*
+// CHECK:   store double [[TMP3]], double* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_f64(float64_t  *a, float64x1_t b) {
   vst1_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst1_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = extractelement <8 x i8> [[B:%.*]], i32 7
-// CHECK-NEXT:    store i8 [[TMP0]], i8* [[A:%.*]], align 1
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_p8(i8* %a, <8 x i8> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = extractelement <8 x i8> %b, i32 7
+// CHECK:   store i8 [[TMP0]], i8* %a
+// CHECK:   ret void
 void test_vst1_lane_p8(poly8_t  *a, poly8x8_t b) {
   vst1_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst1_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
-// CHECK-NEXT:    store i16 [[TMP3]], i16* [[TMP4]], align 2
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_p16(i16* %a, <4 x i16> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i16* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <4 x i16> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// CHECK:   [[TMP3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i16*
+// CHECK:   store i16 [[TMP3]], i16* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_p16(poly16_t  *a, poly16x4_t b) {
   vst1_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst1_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[TMP3:%.*]] = extractelement <1 x i64> [[TMP2]], i32 0
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
-// CHECK-NEXT:    store i64 [[TMP3]], i64* [[TMP4]], align 8
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst1_lane_p64(i64* %a, <1 x i64> %b) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// CHECK:   [[TMP3:%.*]] = extractelement <1 x i64> [[TMP2]], i32 0
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP0]] to i64*
+// CHECK:   store i64 [[TMP3]], i64* [[TMP4]]
+// CHECK:   ret void
 void test_vst1_lane_p64(poly64_t  *a, poly64x1_t b) {
   vst1_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X16X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X16X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <16 x i8>] [[B_COERCE:%.*]], [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x16x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x16x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X2_T]], %struct.uint8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_u8(i8* %a, [2 x <16 x i8>] %b.coerce) #2 {
+// 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.st2lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst2q_lane_u8(uint8_t  *a, uint8x16x2_t b) {
   vst2q_lane_u8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i16>] [[B_COERCE:%.*]], [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X2_T]], %struct.uint16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8i16.p0i8(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_u16(i16* %a, [2 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v8i16.p0i8(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_u16(uint16_t  *a, uint16x8x2_t b) {
   vst2q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X4X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X4X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i32>] [[B_COERCE:%.*]], [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X2_T]], %struct.uint32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4i32.p0i8(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_u32(i32* %a, [2 x <4 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4i32.p0i8(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_u32(uint32_t  *a, uint32x4x2_t b) {
   vst2q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i64>] [[B_COERCE:%.*]], [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X2_T]], %struct.uint64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_u64(i64* %a, [2 x <2 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_u64(uint64_t  *a, uint64x2x2_t b) {
   vst2q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X16X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X16X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <16 x i8>] [[B_COERCE:%.*]], [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x16x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x16x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X2_T]], %struct.int8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_s8(i8* %a, [2 x <16 x i8>] %b.coerce) #2 {
+// 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.st2lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst2q_lane_s8(int8_t  *a, int8x16x2_t b) {
   vst2q_lane_s8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i16>] [[B_COERCE:%.*]], [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X2_T]], %struct.int16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8i16.p0i8(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_s16(i16* %a, [2 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v8i16.p0i8(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_s16(int16_t  *a, int16x8x2_t b) {
   vst2q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X4X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X4X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i32>] [[B_COERCE:%.*]], [2 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X2_T]], %struct.int32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i32>], [2 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4i32.p0i8(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_s32(i32* %a, [2 x <4 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4i32.p0i8(<4 x i32> [[TMP7]], <4 x i32> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_s32(int32_t  *a, int32x4x2_t b) {
   vst2q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i64>] [[B_COERCE:%.*]], [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X2_T]], %struct.int64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_s64(i64* %a, [2 x <2 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_s64(int64_t  *a, int64x2x2_t b) {
   vst2q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x half>] [[B_COERCE:%.*]], [2 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X2_T]], %struct.float16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x half>], [2 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x half>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x half>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8f16.p0i8(<8 x half> [[TMP7]], <8 x half> [[TMP8]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_f16(half* %a, [2 x <8 x half>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v8f16.p0i8(<8 x half> [[TMP7]], <8 x half> [[TMP8]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_f16(float16_t  *a, float16x8x2_t b) {
   vst2q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X4X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X4X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x float>] [[B_COERCE:%.*]], [2 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X2_T]], %struct.float32x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x float>], [2 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4f32.p0i8(<4 x float> [[TMP7]], <4 x float> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_f32(float* %a, [2 x <4 x float>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4f32.p0i8(<4 x float> [[TMP7]], <4 x float> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_f32(float32_t  *a, float32x4x2_t b) {
   vst2q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x double>] [[B_COERCE:%.*]], [2 x <2 x double>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x double> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X2_T]], %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x double> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x double>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2f64.p0i8(<2 x double> [[TMP7]], <2 x double> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_f64(double* %a, [2 x <2 x double>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x2x2_t, %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP4:%.*]] = bitcast <2 x double> [[TMP3]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x2x2_t, %struct.float64x2x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x double>], [2 x <2 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP6:%.*]] = bitcast <2 x double> [[TMP5]] to <16 x i8>
+// CHECK:   [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x double>
+// CHECK:   [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
+// CHECK:   call void @llvm.aarch64.neon.st2lane.v2f64.p0i8(<2 x double> [[TMP7]], <2 x double> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_f64(float64_t  *a, float64x2x2_t b) {
   vst2q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X16X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X16X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <16 x i8>] [[B_COERCE:%.*]], [2 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x16x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x16x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X2_T]], %struct.poly8x16x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <16 x i8>], [2 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_p8(i8* %a, [2 x <16 x i8>] %b.coerce) #2 {
+// 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.st2lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst2q_lane_p8(poly8_t  *a, poly8x16x2_t b) {
   vst2q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X8X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X8X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i16>] [[B_COERCE:%.*]], [2 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X2_T]], %struct.poly16x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i16>], [2 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8i16.p0i8(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_p16(i16* %a, [2 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v8i16.p0i8(<8 x i16> [[TMP7]], <8 x i16> [[TMP8]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_p16(poly16_t  *a, poly16x8x2_t b) {
   vst2q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X2X2_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X2X2_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i64>] [[B_COERCE:%.*]], [2 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X2_T]], %struct.poly64x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i64>], [2 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2q_lane_p64(i64* %a, [2 x <2 x i64>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x2x2_t, %struct.poly64x2x2_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.poly64x2x2_t, %struct.poly64x2x2_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:   call void @llvm.aarch64.neon.st2lane.v2i64.p0i8(<2 x i64> [[TMP7]], <2 x i64> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2q_lane_p64(poly64_t  *a, poly64x2x2_t b) {
   vst2q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X8X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X8X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i8>] [[B_COERCE:%.*]], [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X2_T]], %struct.uint8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_u8(i8* %a, [2 x <8 x i8>] %b.coerce) #2 {
+// 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.st2lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst2_lane_u8(uint8_t  *a, uint8x8x2_t b) {
   vst2_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i16>] [[B_COERCE:%.*]], [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X2_T]], %struct.uint16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4i16.p0i8(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_u16(i16* %a, [2 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4i16.p0i8(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_u16(uint16_t  *a, uint16x4x2_t b) {
   vst2_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X2X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X2X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i32>] [[B_COERCE:%.*]], [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X2_T]], %struct.uint32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2i32.p0i8(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_u32(i32* %a, [2 x <2 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v2i32.p0i8(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_u32(uint32_t  *a, uint32x2x2_t b) {
   vst2_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x i64>] [[B_COERCE:%.*]], [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X2_T]], %struct.uint64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_u64(i64* %a, [2 x <1 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_u64(uint64_t  *a, uint64x1x2_t b) {
   vst2_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst2_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X8X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X8X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i8>] [[B_COERCE:%.*]], [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X2_T]], %struct.int8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_s8(i8* %a, [2 x <8 x i8>] %b.coerce) #2 {
+// 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.st2lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst2_lane_s8(int8_t  *a, int8x8x2_t b) {
   vst2_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i16>] [[B_COERCE:%.*]], [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X2_T]], %struct.int16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4i16.p0i8(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_s16(i16* %a, [2 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4i16.p0i8(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_s16(int16_t  *a, int16x4x2_t b) {
   vst2_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X2X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X2X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x i32>] [[B_COERCE:%.*]], [2 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X2_T]], %struct.int32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x i32>], [2 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2i32.p0i8(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_s32(i32* %a, [2 x <2 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v2i32.p0i8(<2 x i32> [[TMP7]], <2 x i32> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_s32(int32_t  *a, int32x2x2_t b) {
   vst2_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x i64>] [[B_COERCE:%.*]], [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X2_T]], %struct.int64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_s64(i64* %a, [2 x <1 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_s64(int64_t  *a, int64x1x2_t b) {
   vst2_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst2_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x half>] [[B_COERCE:%.*]], [2 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X2_T]], %struct.float16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x half>], [2 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x half>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x half>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4f16.p0i8(<4 x half> [[TMP7]], <4 x half> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_f16(half* %a, [2 x <4 x half>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4f16.p0i8(<4 x half> [[TMP7]], <4 x half> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_f16(float16_t  *a, float16x4x2_t b) {
   vst2_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X2X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X2X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <2 x float>] [[B_COERCE:%.*]], [2 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x2x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x2x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X2_T]], %struct.float32x2x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <2 x float>], [2 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v2f32.p0i8(<2 x float> [[TMP7]], <2 x float> [[TMP8]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_f32(float* %a, [2 x <2 x float>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v2f32.p0i8(<2 x float> [[TMP7]], <2 x float> [[TMP8]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_f32(float32_t  *a, float32x2x2_t b) {
   vst2_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst2_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x double>] [[B_COERCE:%.*]], [2 x <1 x double>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x double> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X2_T]], %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x double> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x double>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v1f64.p0i8(<1 x double> [[TMP7]], <1 x double> [[TMP8]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_f64(double* %a, [2 x <1 x double>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x1x2_t, %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP3:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP4:%.*]] = bitcast <1 x double> [[TMP3]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x1x2_t, %struct.float64x1x2_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x double>], [2 x <1 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP5:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP6:%.*]] = bitcast <1 x double> [[TMP5]] to <8 x i8>
+// CHECK:   [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x double>
+// CHECK:   [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
+// CHECK:   call void @llvm.aarch64.neon.st2lane.v1f64.p0i8(<1 x double> [[TMP7]], <1 x double> [[TMP8]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_f64(float64_t  *a, float64x1x2_t b) {
   vst2_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst2_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X8X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X8X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <8 x i8>] [[B_COERCE:%.*]], [2 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x8x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x8x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X2_T]], %struct.poly8x8x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <8 x i8>], [2 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_p8(i8* %a, [2 x <8 x i8>] %b.coerce) #2 {
+// 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.st2lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst2_lane_p8(poly8_t  *a, poly8x8x2_t b) {
   vst2_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst2_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X4X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X4X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <4 x i16>] [[B_COERCE:%.*]], [2 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x4x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x4x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X2_T]], %struct.poly16x4x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <4 x i16>], [2 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v4i16.p0i8(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_p16(i16* %a, [2 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st2lane.v4i16.p0i8(<4 x i16> [[TMP7]], <4 x i16> [[TMP8]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_p16(poly16_t  *a, poly16x4x2_t b) {
   vst2_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst2_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X1X2_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X1X2_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [2 x <1 x i64>] [[B_COERCE:%.*]], [2 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x2_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x1x2_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 16, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X2_T]], %struct.poly64x1x2_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x <1 x i64>], [2 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st2lane.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst2_lane_p64(i64* %a, [2 x <1 x i64>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x1x2_t, %struct.poly64x1x2_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.poly64x1x2_t, %struct.poly64x1x2_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:   call void @llvm.aarch64.neon.st2lane.v1i64.p0i8(<1 x i64> [[TMP7]], <1 x i64> [[TMP8]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst2_lane_p64(poly64_t  *a, poly64x1x2_t b) {
   vst2_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X16X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X16X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <16 x i8>] [[B_COERCE:%.*]], [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x16x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x16x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X3_T]], %struct.uint8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_u8(i8* %a, [3 x <16 x i8>] %b.coerce) #2 {
+// 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.st3lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst3q_lane_u8(uint8_t  *a, uint8x16x3_t b) {
   vst3q_lane_u8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i16>] [[B_COERCE:%.*]], [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X3_T]], %struct.uint16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8i16.p0i8(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_u16(i16* %a, [3 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v8i16.p0i8(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_u16(uint16_t  *a, uint16x8x3_t b) {
   vst3q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X4X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X4X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i32>] [[B_COERCE:%.*]], [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X3_T]], %struct.uint32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4i32.p0i8(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_u32(i32* %a, [3 x <4 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4i32.p0i8(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_u32(uint32_t  *a, uint32x4x3_t b) {
   vst3q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i64>] [[B_COERCE:%.*]], [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X3_T]], %struct.uint64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_u64(i64* %a, [3 x <2 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_u64(uint64_t  *a, uint64x2x3_t b) {
   vst3q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X16X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X16X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <16 x i8>] [[B_COERCE:%.*]], [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x16x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x16x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X3_T]], %struct.int8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_s8(i8* %a, [3 x <16 x i8>] %b.coerce) #2 {
+// 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.st3lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst3q_lane_s8(int8_t  *a, int8x16x3_t b) {
   vst3q_lane_s8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i16>] [[B_COERCE:%.*]], [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X3_T]], %struct.int16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8i16.p0i8(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_s16(i16* %a, [3 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v8i16.p0i8(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_s16(int16_t  *a, int16x8x3_t b) {
   vst3q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X4X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X4X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i32>] [[B_COERCE:%.*]], [3 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X3_T]], %struct.int32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4i32.p0i8(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_s32(i32* %a, [3 x <4 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4i32.p0i8(<4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <4 x i32> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_s32(int32_t  *a, int32x4x3_t b) {
   vst3q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i64>] [[B_COERCE:%.*]], [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X3_T]], %struct.int64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_s64(i64* %a, [3 x <2 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_s64(int64_t  *a, int64x2x3_t b) {
   vst3q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x half>] [[B_COERCE:%.*]], [3 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X3_T]], %struct.float16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x half>], [3 x <8 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x half> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x half>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x half>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x half>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8f16.p0i8(<8 x half> [[TMP9]], <8 x half> [[TMP10]], <8 x half> [[TMP11]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_f16(half* %a, [3 x <8 x half>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v8f16.p0i8(<8 x half> [[TMP9]], <8 x half> [[TMP10]], <8 x half> [[TMP11]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_f16(float16_t  *a, float16x8x3_t b) {
   vst3q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X4X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X4X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x float>] [[B_COERCE:%.*]], [3 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X3_T]], %struct.float32x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x float>], [3 x <4 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x float> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x float>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4f32.p0i8(<4 x float> [[TMP9]], <4 x float> [[TMP10]], <4 x float> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_f32(float* %a, [3 x <4 x float>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4f32.p0i8(<4 x float> [[TMP9]], <4 x float> [[TMP10]], <4 x float> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_f32(float32_t  *a, float32x4x3_t b) {
   vst3q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x double>] [[B_COERCE:%.*]], [3 x <2 x double>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x double> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x double> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X3_T]], %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x double> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x double>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x double>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2f64.p0i8(<2 x double> [[TMP9]], <2 x double> [[TMP10]], <2 x double> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_f64(double* %a, [3 x <2 x double>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP4:%.*]] = bitcast <2 x double> [[TMP3]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP6:%.*]] = bitcast <2 x double> [[TMP5]] to <16 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x2x3_t, %struct.float64x2x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x double>], [3 x <2 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP7:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP8:%.*]] = bitcast <2 x double> [[TMP7]] to <16 x i8>
+// CHECK:   [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x double>
+// CHECK:   [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x double>
+// CHECK:   call void @llvm.aarch64.neon.st3lane.v2f64.p0i8(<2 x double> [[TMP9]], <2 x double> [[TMP10]], <2 x double> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_f64(float64_t  *a, float64x2x3_t b) {
   vst3q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X16X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X16X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <16 x i8>] [[B_COERCE:%.*]], [3 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x16x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x16x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X3_T]], %struct.poly8x16x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <16 x i8>], [3 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_p8(i8* %a, [3 x <16 x i8>] %b.coerce) #2 {
+// 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.st3lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst3q_lane_p8(poly8_t  *a, poly8x16x3_t b) {
   vst3q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X8X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X8X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i16>] [[B_COERCE:%.*]], [3 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X3_T]], %struct.poly16x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i16>], [3 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8i16.p0i8(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_p16(i16* %a, [3 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v8i16.p0i8(<8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <8 x i16> [[TMP11]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_p16(poly16_t  *a, poly16x8x3_t b) {
   vst3q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X2X3_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X2X3_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i64>] [[B_COERCE:%.*]], [3 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 48, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X3_T]], %struct.poly64x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i64>], [3 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3q_lane_p64(i64* %a, [3 x <2 x i64>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x2x3_t, %struct.poly64x2x3_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.poly64x2x3_t, %struct.poly64x2x3_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.poly64x2x3_t, %struct.poly64x2x3_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:   call void @llvm.aarch64.neon.st3lane.v2i64.p0i8(<2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <2 x i64> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3q_lane_p64(poly64_t  *a, poly64x2x3_t b) {
   vst3q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X8X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X8X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i8>] [[B_COERCE:%.*]], [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X3_T]], %struct.uint8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_u8(i8* %a, [3 x <8 x i8>] %b.coerce) #2 {
+// 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.st3lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst3_lane_u8(uint8_t  *a, uint8x8x3_t b) {
   vst3_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i16>] [[B_COERCE:%.*]], [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X3_T]], %struct.uint16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4i16.p0i8(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_u16(i16* %a, [3 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4i16.p0i8(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_u16(uint16_t  *a, uint16x4x3_t b) {
   vst3_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X2X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X2X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i32>] [[B_COERCE:%.*]], [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X3_T]], %struct.uint32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2i32.p0i8(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_u32(i32* %a, [3 x <2 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v2i32.p0i8(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_u32(uint32_t  *a, uint32x2x3_t b) {
   vst3_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x i64>] [[B_COERCE:%.*]], [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X3_T]], %struct.uint64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_u64(i64* %a, [3 x <1 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_u64(uint64_t  *a, uint64x1x3_t b) {
   vst3_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst3_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X8X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X8X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i8>] [[B_COERCE:%.*]], [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X3_T]], %struct.int8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_s8(i8* %a, [3 x <8 x i8>] %b.coerce) #2 {
+// 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.st3lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst3_lane_s8(int8_t  *a, int8x8x3_t b) {
   vst3_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i16>] [[B_COERCE:%.*]], [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X3_T]], %struct.int16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4i16.p0i8(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_s16(i16* %a, [3 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4i16.p0i8(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_s16(int16_t  *a, int16x4x3_t b) {
   vst3_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X2X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X2X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x i32>] [[B_COERCE:%.*]], [3 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X3_T]], %struct.int32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x i32>], [3 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2i32.p0i8(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_s32(i32* %a, [3 x <2 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v2i32.p0i8(<2 x i32> [[TMP9]], <2 x i32> [[TMP10]], <2 x i32> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_s32(int32_t  *a, int32x2x3_t b) {
   vst3_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x i64>] [[B_COERCE:%.*]], [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X3_T]], %struct.int64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_s64(i64* %a, [3 x <1 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_s64(int64_t  *a, int64x1x3_t b) {
   vst3_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst3_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x half>] [[B_COERCE:%.*]], [3 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X3_T]], %struct.float16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x half>], [3 x <4 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x half> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x half>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x half>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x half>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4f16.p0i8(<4 x half> [[TMP9]], <4 x half> [[TMP10]], <4 x half> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_f16(half* %a, [3 x <4 x half>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4f16.p0i8(<4 x half> [[TMP9]], <4 x half> [[TMP10]], <4 x half> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_f16(float16_t  *a, float16x4x3_t b) {
   vst3_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X2X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X2X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <2 x float>] [[B_COERCE:%.*]], [3 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x2x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x2x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X3_T]], %struct.float32x2x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <2 x float>], [3 x <2 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x float>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v2f32.p0i8(<2 x float> [[TMP9]], <2 x float> [[TMP10]], <2 x float> [[TMP11]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_f32(float* %a, [3 x <2 x float>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v2f32.p0i8(<2 x float> [[TMP9]], <2 x float> [[TMP10]], <2 x float> [[TMP11]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_f32(float32_t  *a, float32x2x3_t b) {
   vst3_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst3_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x double>] [[B_COERCE:%.*]], [3 x <1 x double>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x double> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x double> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X3_T]], %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x double> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x double>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x double>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v1f64.p0i8(<1 x double> [[TMP9]], <1 x double> [[TMP10]], <1 x double> [[TMP11]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_f64(double* %a, [3 x <1 x double>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP3:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP4:%.*]] = bitcast <1 x double> [[TMP3]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP5:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP6:%.*]] = bitcast <1 x double> [[TMP5]] to <8 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x1x3_t, %struct.float64x1x3_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x double>], [3 x <1 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP7:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP8:%.*]] = bitcast <1 x double> [[TMP7]] to <8 x i8>
+// CHECK:   [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x double>
+// CHECK:   [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x double>
+// CHECK:   call void @llvm.aarch64.neon.st3lane.v1f64.p0i8(<1 x double> [[TMP9]], <1 x double> [[TMP10]], <1 x double> [[TMP11]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_f64(float64_t  *a, float64x1x3_t b) {
   vst3_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst3_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X8X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X8X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <8 x i8>] [[B_COERCE:%.*]], [3 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x8x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x8x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X3_T]], %struct.poly8x8x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <8 x i8>], [3 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_p8(i8* %a, [3 x <8 x i8>] %b.coerce) #2 {
+// 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.st3lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst3_lane_p8(poly8_t  *a, poly8x8x3_t b) {
   vst3_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst3_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X4X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X4X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <4 x i16>] [[B_COERCE:%.*]], [3 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x4x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x4x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X3_T]], %struct.poly16x4x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <4 x i16>], [3 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v4i16.p0i8(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_p16(i16* %a, [3 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st3lane.v4i16.p0i8(<4 x i16> [[TMP9]], <4 x i16> [[TMP10]], <4 x i16> [[TMP11]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_p16(poly16_t  *a, poly16x4x3_t b) {
   vst3_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst3_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X1X3_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X1X3_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [3 x <1 x i64>] [[B_COERCE:%.*]], [3 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x3_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x1x3_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 24, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X3_T]], %struct.poly64x1x3_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x <1 x i64>], [3 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st3lane.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst3_lane_p64(i64* %a, [3 x <1 x i64>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x1x3_t, %struct.poly64x1x3_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.poly64x1x3_t, %struct.poly64x1x3_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.poly64x1x3_t, %struct.poly64x1x3_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:   call void @llvm.aarch64.neon.st3lane.v1i64.p0i8(<1 x i64> [[TMP9]], <1 x i64> [[TMP10]], <1 x i64> [[TMP11]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst3_lane_p64(poly64_t  *a, poly64x1x3_t b) {
   vst3_lane_p64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X16X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X16X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <16 x i8>] [[B_COERCE:%.*]], [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x16x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x16x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT8X16X4_T]], %struct.uint8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_u8(i8* %a, [4 x <16 x i8>] %b.coerce) #2 {
+// 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.st4lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst4q_lane_u8(uint8_t  *a, uint8x16x4_t b) {
   vst4q_lane_u8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i16>] [[B_COERCE:%.*]], [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT16X8X4_T]], %struct.uint16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8i16.p0i8(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_u16(i16* %a, [4 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v8i16.p0i8(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_u16(uint16_t  *a, uint16x8x4_t b) {
   vst4q_lane_u16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X4X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X4X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i32>] [[B_COERCE:%.*]], [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT32X4X4_T]], %struct.uint32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x i32> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4i32.p0i8(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_u32(i32* %a, [4 x <4 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4i32.p0i8(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_u32(uint32_t  *a, uint32x4x4_t b) {
   vst4q_lane_u32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i64>] [[B_COERCE:%.*]], [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT64X2X4_T]], %struct.uint64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_u64(i64* %a, [4 x <2 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_u64(uint64_t  *a, uint64x2x4_t b) {
   vst4q_lane_u64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X16X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X16X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <16 x i8>] [[B_COERCE:%.*]], [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x16x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x16x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT8X16X4_T]], %struct.int8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_s8(i8* %a, [4 x <16 x i8>] %b.coerce) #2 {
+// 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.st4lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst4q_lane_s8(int8_t  *a, int8x16x4_t b) {
   vst4q_lane_s8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i16>] [[B_COERCE:%.*]], [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT16X8X4_T]], %struct.int16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8i16.p0i8(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_s16(i16* %a, [4 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v8i16.p0i8(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_s16(int16_t  *a, int16x8x4_t b) {
   vst4q_lane_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X4X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X4X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i32>] [[B_COERCE:%.*]], [4 x <4 x i32>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i32> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i32> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i32> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT32X4X4_T]], %struct.int32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i32>], [4 x <4 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x i32> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4i32.p0i8(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_s32(i32* %a, [4 x <4 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4i32.p0i8(<4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <4 x i32> [[TMP13]], <4 x i32> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_s32(int32_t  *a, int32x4x4_t b) {
   vst4q_lane_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i64>] [[B_COERCE:%.*]], [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT64X2X4_T]], %struct.int64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_s64(i64* %a, [4 x <2 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_s64(int64_t  *a, int64x2x4_t b) {
   vst4q_lane_s64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x half>] [[B_COERCE:%.*]], [4 x <8 x half>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x half> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x half> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x half> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X8X4_T]], %struct.float16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x half>], [4 x <8 x half>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <8 x half>, <8 x half>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x half> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x half>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x half>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x half>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x half>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8f16.p0i8(<8 x half> [[TMP11]], <8 x half> [[TMP12]], <8 x half> [[TMP13]], <8 x half> [[TMP14]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_f16(half* %a, [4 x <8 x half>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v8f16.p0i8(<8 x half> [[TMP11]], <8 x half> [[TMP12]], <8 x half> [[TMP13]], <8 x half> [[TMP14]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_f16(float16_t  *a, float16x8x4_t b) {
   vst4q_lane_f16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X4X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X4X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x float>] [[B_COERCE:%.*]], [4 x <4 x float>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x float> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x float> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x float> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X4X4_T]], %struct.float32x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x float>], [4 x <4 x float>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x float>, <4 x float>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x float> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <4 x float>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <4 x float>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <4 x float>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <4 x float>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4f32.p0i8(<4 x float> [[TMP11]], <4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_f32(float* %a, [4 x <4 x float>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4f32.p0i8(<4 x float> [[TMP11]], <4 x float> [[TMP12]], <4 x float> [[TMP13]], <4 x float> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_f32(float32_t  *a, float32x4x4_t b) {
   vst4q_lane_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x double>] [[B_COERCE:%.*]], [4 x <2 x double>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x double> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x double> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x double> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X2X4_T]], %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x double> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x double>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x double>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x double>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2f64.p0i8(<2 x double> [[TMP11]], <2 x double> [[TMP12]], <2 x double> [[TMP13]], <2 x double> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_f64(double* %a, [4 x <2 x double>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX]], align 16
+// CHECK:   [[TMP4:%.*]] = bitcast <2 x double> [[TMP3]] to <16 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX2]], align 16
+// CHECK:   [[TMP6:%.*]] = bitcast <2 x double> [[TMP5]] to <16 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP7:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX4]], align 16
+// CHECK:   [[TMP8:%.*]] = bitcast <2 x double> [[TMP7]] to <16 x i8>
+// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float64x2x4_t, %struct.float64x2x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x double>], [4 x <2 x double>]* [[VAL5]], i64 0, i64 3
+// CHECK:   [[TMP9:%.*]] = load <2 x double>, <2 x double>* [[ARRAYIDX6]], align 16
+// CHECK:   [[TMP10:%.*]] = bitcast <2 x double> [[TMP9]] to <16 x i8>
+// CHECK:   [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x double>
+// CHECK:   [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x double>
+// CHECK:   [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x double>
+// CHECK:   [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x double>
+// CHECK:   call void @llvm.aarch64.neon.st4lane.v2f64.p0i8(<2 x double> [[TMP11]], <2 x double> [[TMP12]], <2 x double> [[TMP13]], <2 x double> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_f64(float64_t  *a, float64x2x4_t b) {
   vst4q_lane_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X16X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X16X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <16 x i8>] [[B_COERCE:%.*]], [4 x <16 x i8>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x16x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x16x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY8X16X4_T]], %struct.poly8x16x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <16 x i8>], [4 x <16 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_p8(i8* %a, [4 x <16 x i8>] %b.coerce) #2 {
+// 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.st4lane.v16i8.p0i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP3]], <16 x i8> [[TMP4]], <16 x i8> [[TMP5]], i64 15, i8* %a)
+// CHECK:   ret void
 void test_vst4q_lane_p8(poly8_t  *a, poly8x16x4_t b) {
   vst4q_lane_p8(a, b, 15);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X8X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X8X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i16>] [[B_COERCE:%.*]], [4 x <8 x i16>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i16> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <8 x i16> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <8 x i16> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY16X8X4_T]], %struct.poly16x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i16>], [4 x <8 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <8 x i16>, <8 x i16>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <8 x i16> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <8 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <8 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <8 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <8 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8i16.p0i8(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i64 7, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_p16(i16* %a, [4 x <8 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v8i16.p0i8(<8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <8 x i16> [[TMP13]], <8 x i16> [[TMP14]], i64 7, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_p16(poly16_t  *a, poly16x8x4_t b) {
   vst4q_lane_p16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4q_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X2X4_T:%.*]], align 16
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X2X4_T]], align 16
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i64>] [[B_COERCE:%.*]], [4 x <2 x i64>]* [[COERCE_DIVE]], align 16
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 16 [[TMP0]], i8* align 16 [[TMP1]], i64 64, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX]], align 16
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i64> [[TMP3]] to <16 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX2]], align 16
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i64> [[TMP5]] to <16 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX4]], align 16
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i64> [[TMP7]] to <16 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY64X2X4_T]], %struct.poly64x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i64>], [4 x <2 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x i64>, <2 x i64>* [[ARRAYIDX6]], align 16
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x i64> [[TMP9]] to <16 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <16 x i8> [[TMP4]] to <2 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <16 x i8> [[TMP6]] to <2 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <16 x i8> [[TMP8]] to <2 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <16 x i8> [[TMP10]] to <2 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4q_lane_p64(i64* %a, [4 x <2 x i64>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x2x4_t, %struct.poly64x2x4_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.poly64x2x4_t, %struct.poly64x2x4_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.poly64x2x4_t, %struct.poly64x2x4_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.poly64x2x4_t, %struct.poly64x2x4_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:   call void @llvm.aarch64.neon.st4lane.v2i64.p0i8(<2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <2 x i64> [[TMP13]], <2 x i64> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4q_lane_p64(poly64_t  *a, poly64x2x4_t b) {
   vst4q_lane_p64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4_lane_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT8X8X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT8X8X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i8>] [[B_COERCE:%.*]], [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint8x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint8x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT8X8X4_T]], %struct.uint8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_u8(i8* %a, [4 x <8 x i8>] %b.coerce) #2 {
+// 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.st4lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst4_lane_u8(uint8_t  *a, uint8x8x4_t b) {
   vst4_lane_u8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4_lane_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i16>] [[B_COERCE:%.*]], [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT16X4X4_T]], %struct.uint16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4i16.p0i8(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_u16(i16* %a, [4 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4i16.p0i8(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_u16(uint16_t  *a, uint16x4x4_t b) {
   vst4_lane_u16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4_lane_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT32X2X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT32X2X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i32>] [[B_COERCE:%.*]], [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint32x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint32x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT32X2X4_T]], %struct.uint32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x i32> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2i32.p0i8(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_u32(i32* %a, [4 x <2 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v2i32.p0i8(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_u32(uint32_t  *a, uint32x2x4_t b) {
   vst4_lane_u32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4_lane_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_UINT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_UINT64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x i64>] [[B_COERCE:%.*]], [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.uint64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.uint64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_UINT64X1X4_T]], %struct.uint64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_u64(i64* %a, [4 x <1 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_u64(uint64_t  *a, uint64x1x4_t b) {
   vst4_lane_u64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst4_lane_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT8X8X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT8X8X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i8>] [[B_COERCE:%.*]], [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int8x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int8x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT8X8X4_T]], %struct.int8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_s8(i8* %a, [4 x <8 x i8>] %b.coerce) #2 {
+// 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.st4lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst4_lane_s8(int8_t  *a, int8x8x4_t b) {
   vst4_lane_s8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i16>] [[B_COERCE:%.*]], [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT16X4X4_T]], %struct.int16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4i16.p0i8(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_s16(i16* %a, [4 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4i16.p0i8(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_s16(int16_t  *a, int16x4x4_t b) {
   vst4_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT32X2X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT32X2X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x i32>] [[B_COERCE:%.*]], [4 x <2 x i32>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int32x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int32x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x i32> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x i32> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x i32> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT32X2X4_T]], %struct.int32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x i32>], [4 x <2 x i32>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x i32> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x i32>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x i32>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x i32>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x i32>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2i32.p0i8(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_s32(i32* %a, [4 x <2 x i32>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v2i32.p0i8(<2 x i32> [[TMP11]], <2 x i32> [[TMP12]], <2 x i32> [[TMP13]], <2 x i32> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_s32(int32_t  *a, int32x2x4_t b) {
   vst4_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4_lane_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_INT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_INT64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x i64>] [[B_COERCE:%.*]], [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.int64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.int64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_INT64X1X4_T]], %struct.int64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_s64(i64* %a, [4 x <1 x i64>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_s64(int64_t  *a, int64x1x4_t b) {
   vst4_lane_s64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst4_lane_f16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x half>] [[B_COERCE:%.*]], [4 x <4 x half>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast half* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x half> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x half> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x half> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT16X4X4_T]], %struct.float16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x half>], [4 x <4 x half>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x half>, <4 x half>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x half> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x half>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x half>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x half>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x half>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4f16.p0i8(<4 x half> [[TMP11]], <4 x half> [[TMP12]], <4 x half> [[TMP13]], <4 x half> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_f16(half* %a, [4 x <4 x half>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4f16.p0i8(<4 x half> [[TMP11]], <4 x half> [[TMP12]], <4 x half> [[TMP13]], <4 x half> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_f16(float16_t  *a, float16x4x4_t b) {
   vst4_lane_f16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT32X2X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT32X2X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <2 x float>] [[B_COERCE:%.*]], [4 x <2 x float>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float32x2x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float32x2x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <2 x float> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <2 x float> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <2 x float> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT32X2X4_T]], %struct.float32x2x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <2 x float>], [4 x <2 x float>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <2 x float>, <2 x float>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <2 x float> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <2 x float>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <2 x float>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <2 x float>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <2 x float>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v2f32.p0i8(<2 x float> [[TMP11]], <2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]], i64 1, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_f32(float* %a, [4 x <2 x float>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v2f32.p0i8(<2 x float> [[TMP11]], <2 x float> [[TMP12]], <2 x float> [[TMP13]], <2 x float> [[TMP14]], i64 1, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_f32(float32_t  *a, float32x2x4_t b) {
   vst4_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vst4_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_FLOAT64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x double>] [[B_COERCE:%.*]], [4 x <1 x double>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.float64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.float64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x double> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x double> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x double> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_FLOAT64X1X4_T]], %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <1 x double> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x double>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x double>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x double>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v1f64.p0i8(<1 x double> [[TMP11]], <1 x double> [[TMP12]], <1 x double> [[TMP13]], <1 x double> [[TMP14]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_f64(double* %a, [4 x <1 x double>] %b.coerce) #2 {
+// 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:   [[TMP2:%.*]] = bitcast double* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL]], i64 0, i64 0
+// CHECK:   [[TMP3:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX]], align 8
+// CHECK:   [[TMP4:%.*]] = bitcast <1 x double> [[TMP3]] to <8 x i8>
+// CHECK:   [[VAL1:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL1]], i64 0, i64 1
+// CHECK:   [[TMP5:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX2]], align 8
+// CHECK:   [[TMP6:%.*]] = bitcast <1 x double> [[TMP5]] to <8 x i8>
+// CHECK:   [[VAL3:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL3]], i64 0, i64 2
+// CHECK:   [[TMP7:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX4]], align 8
+// CHECK:   [[TMP8:%.*]] = bitcast <1 x double> [[TMP7]] to <8 x i8>
+// CHECK:   [[VAL5:%.*]] = getelementptr inbounds %struct.float64x1x4_t, %struct.float64x1x4_t* [[__S1]], i32 0, i32 0
+// CHECK:   [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x double>], [4 x <1 x double>]* [[VAL5]], i64 0, i64 3
+// CHECK:   [[TMP9:%.*]] = load <1 x double>, <1 x double>* [[ARRAYIDX6]], align 8
+// CHECK:   [[TMP10:%.*]] = bitcast <1 x double> [[TMP9]] to <8 x i8>
+// CHECK:   [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x double>
+// CHECK:   [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x double>
+// CHECK:   [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x double>
+// CHECK:   [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x double>
+// CHECK:   call void @llvm.aarch64.neon.st4lane.v1f64.p0i8(<1 x double> [[TMP11]], <1 x double> [[TMP12]], <1 x double> [[TMP13]], <1 x double> [[TMP14]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_f64(float64_t  *a, float64x1x4_t b) {
   vst4_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vst4_lane_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY8X8X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY8X8X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <8 x i8>] [[B_COERCE:%.*]], [4 x <8 x i8>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly8x8x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly8x8x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY8X8X4_T]], %struct.poly8x8x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <8 x i8>], [4 x <8 x i8>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP5:%.*]] = load <8 x i8>, <8 x i8>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* [[A:%.*]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_p8(i8* %a, [4 x <8 x i8>] %b.coerce) #2 {
+// 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.st4lane.v8i8.p0i8(<8 x i8> [[TMP2]], <8 x i8> [[TMP3]], <8 x i8> [[TMP4]], <8 x i8> [[TMP5]], i64 7, i8* %a)
+// CHECK:   ret void
 void test_vst4_lane_p8(poly8_t  *a, poly8x8x4_t b) {
   vst4_lane_p8(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vst4_lane_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY16X4X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY16X4X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <4 x i16>] [[B_COERCE:%.*]], [4 x <4 x i16>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly16x4x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly16x4x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i16* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <4 x i16> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <4 x i16> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY16X4X4_T]], %struct.poly16x4x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <4 x i16>], [4 x <4 x i16>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i16>, <4 x i16>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <4 x i16> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <4 x i16>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <4 x i16>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <4 x i16>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <4 x i16>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v4i16.p0i8(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i64 3, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_p16(i16* %a, [4 x <4 x i16>] %b.coerce) #2 {
+// 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:   call void @llvm.aarch64.neon.st4lane.v4i16.p0i8(<4 x i16> [[TMP11]], <4 x i16> [[TMP12]], <4 x i16> [[TMP13]], <4 x i16> [[TMP14]], i64 3, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_p16(poly16_t  *a, poly16x4x4_t b) {
   vst4_lane_p16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vst4_lane_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[B:%.*]] = alloca [[STRUCT_POLY64X1X4_T:%.*]], align 8
-// CHECK-NEXT:    [[__S1:%.*]] = alloca [[STRUCT_POLY64X1X4_T]], align 8
-// CHECK-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[B]], i32 0, i32 0
-// CHECK-NEXT:    store [4 x <1 x i64>] [[B_COERCE:%.*]], [4 x <1 x i64>]* [[COERCE_DIVE]], align 8
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast %struct.poly64x1x4_t* [[__S1]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast %struct.poly64x1x4_t* [[B]] to i8*
-// CHECK-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP0]], i8* align 8 [[TMP1]], i64 32, i1 false)
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast i64* [[A:%.*]] to i8*
-// CHECK-NEXT:    [[VAL:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL]], i64 0, i64 0
-// CHECK-NEXT:    [[TMP3:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX]], align 8
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <1 x i64> [[TMP3]] to <8 x i8>
-// CHECK-NEXT:    [[VAL1:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL1]], i64 0, i64 1
-// CHECK-NEXT:    [[TMP5:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX2]], align 8
-// CHECK-NEXT:    [[TMP6:%.*]] = bitcast <1 x i64> [[TMP5]] to <8 x i8>
-// CHECK-NEXT:    [[VAL3:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL3]], i64 0, i64 2
-// CHECK-NEXT:    [[TMP7:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX4]], align 8
-// CHECK-NEXT:    [[TMP8:%.*]] = bitcast <1 x i64> [[TMP7]] to <8 x i8>
-// CHECK-NEXT:    [[VAL5:%.*]] = getelementptr inbounds [[STRUCT_POLY64X1X4_T]], %struct.poly64x1x4_t* [[__S1]], i32 0, i32 0
-// CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [4 x <1 x i64>], [4 x <1 x i64>]* [[VAL5]], i64 0, i64 3
-// CHECK-NEXT:    [[TMP9:%.*]] = load <1 x i64>, <1 x i64>* [[ARRAYIDX6]], align 8
-// CHECK-NEXT:    [[TMP10:%.*]] = bitcast <1 x i64> [[TMP9]] to <8 x i8>
-// CHECK-NEXT:    [[TMP11:%.*]] = bitcast <8 x i8> [[TMP4]] to <1 x i64>
-// CHECK-NEXT:    [[TMP12:%.*]] = bitcast <8 x i8> [[TMP6]] to <1 x i64>
-// CHECK-NEXT:    [[TMP13:%.*]] = bitcast <8 x i8> [[TMP8]] to <1 x i64>
-// CHECK-NEXT:    [[TMP14:%.*]] = bitcast <8 x i8> [[TMP10]] to <1 x i64>
-// CHECK-NEXT:    call void @llvm.aarch64.neon.st4lane.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64 0, i8* [[TMP2]])
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vst4_lane_p64(i64* %a, [4 x <1 x i64>] %b.coerce) #2 {
+// 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)
+// CHECK:   [[TMP2:%.*]] = bitcast i64* %a to i8*
+// CHECK:   [[VAL:%.*]] = getelementptr inbounds %struct.poly64x1x4_t, %struct.poly64x1x4_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.poly64x1x4_t, %struct.poly64x1x4_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.poly64x1x4_t, %struct.poly64x1x4_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.poly64x1x4_t, %struct.poly64x1x4_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:   call void @llvm.aarch64.neon.st4lane.v1i64.p0i8(<1 x i64> [[TMP11]], <1 x i64> [[TMP12]], <1 x i64> [[TMP13]], <1 x i64> [[TMP14]], i64 0, i8* [[TMP2]])
+// CHECK:   ret void
 void test_vst4_lane_p64(poly64_t  *a, poly64x1x4_t b) {
   vst4_lane_p64(a, b, 0);
 }
 
+// CHECK: attributes #0 ={{.*}}"min-legal-vector-width"="128"
+// CHECK: attributes #1 ={{.*}}"min-legal-vector-width"="64"
+// CHECK: attributes #2 ={{.*}}"min-legal-vector-width"="0"
diff --git a/clang/test/CodeGen/aarch64-neon-scalar-x-indexed-elem.c b/clang/test/CodeGen/aarch64-neon-scalar-x-indexed-elem.c
index 5c9a5c13dc2b..367978d84af4 100644
--- a/clang/test/CodeGen/aarch64-neon-scalar-x-indexed-elem.c
+++ b/clang/test/CodeGen/aarch64-neon-scalar-x-indexed-elem.c
@@ -1,4 +1,3 @@
-// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon -target-cpu cyclone \
 // RUN: -disable-O0-optnone -emit-llvm -o - %s | opt -S -mem2reg | FileCheck %s
 
@@ -7,496 +6,414 @@
 #include <arm_neon.h>
 
 
-// CHECK-LABEL: @test_vmuls_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <2 x float> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[MUL:%.*]] = fmul float [[A:%.*]], [[VGET_LANE]]
-// CHECK-NEXT:    ret float [[MUL]]
-//
+// CHECK-LABEL: define{{.*}} float @test_vmuls_lane_f32(float %a, <2 x float> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <2 x float> %b, i32 1
+// CHECK:   [[MUL:%.*]] = fmul float %a, [[VGET_LANE]]
+// CHECK:   ret float [[MUL]]
 float32_t test_vmuls_lane_f32(float32_t a, float32x2_t b) {
   return vmuls_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vmuld_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[B:%.*]], i32 0
-// CHECK-NEXT:    [[MUL:%.*]] = fmul double [[A:%.*]], [[VGET_LANE]]
-// CHECK-NEXT:    ret double [[MUL]]
-//
+// CHECK-LABEL: define{{.*}} double @test_vmuld_lane_f64(double %a, <1 x double> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> %b, i32 0
+// CHECK:   [[MUL:%.*]] = fmul double %a, [[VGET_LANE]]
+// CHECK:   ret double [[MUL]]
 float64_t test_vmuld_lane_f64(float64_t a, float64x1_t b) {
   return vmuld_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vmuls_laneq_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <4 x float> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[MUL:%.*]] = fmul float [[A:%.*]], [[VGETQ_LANE]]
-// CHECK-NEXT:    ret float [[MUL]]
-//
+// CHECK-LABEL: define{{.*}} float @test_vmuls_laneq_f32(float %a, <4 x float> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <4 x float> %b, i32 3
+// CHECK:   [[MUL:%.*]] = fmul float %a, [[VGETQ_LANE]]
+// CHECK:   ret float [[MUL]]
 float32_t test_vmuls_laneq_f32(float32_t a, float32x4_t b) {
   return vmuls_laneq_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vmuld_laneq_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[MUL:%.*]] = fmul double [[A:%.*]], [[VGETQ_LANE]]
-// CHECK-NEXT:    ret double [[MUL]]
-//
+// CHECK-LABEL: define{{.*}} double @test_vmuld_laneq_f64(double %a, <2 x double> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x double> %b, i32 1
+// CHECK:   [[MUL:%.*]] = fmul double %a, [[VGETQ_LANE]]
+// CHECK:   ret double [[MUL]]
 float64_t test_vmuld_laneq_f64(float64_t a, float64x2_t b) {
   return vmuld_laneq_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vmul_n_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to double
-// CHECK-NEXT:    [[TMP1:%.*]] = fmul double [[TMP0]], [[B:%.*]]
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast double [[TMP1]] to <1 x double>
-// CHECK-NEXT:    ret <1 x double> [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vmul_n_f64(<1 x double> %a, double %b) #0 {
+// CHECK:   [[TMP2:%.*]] = bitcast <1 x double> %a to double
+// CHECK:   [[TMP3:%.*]] = fmul double [[TMP2]], %b
+// CHECK:   [[TMP4:%.*]] = bitcast double [[TMP3]] to <1 x double>
+// CHECK:   ret <1 x double> [[TMP4]]
 float64x1_t test_vmul_n_f64(float64x1_t a, float64_t b) {
   return vmul_n_f64(a, b);
 }
 
-// CHECK-LABEL: @test_vmulxs_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <2 x float> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[VMULXS_F32_I:%.*]] = call float @llvm.aarch64.neon.fmulx.f32(float [[A:%.*]], float [[VGET_LANE]]) [[ATTR5:#.*]]
-// CHECK-NEXT:    ret float [[VMULXS_F32_I]]
-//
+// CHECK-LABEL: define{{.*}} float @test_vmulxs_lane_f32(float %a, <2 x float> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <2 x float> %b, i32 1
+// CHECK:   [[VMULXS_F32_I:%.*]] = call float @llvm.aarch64.neon.fmulx.f32(float %a, float [[VGET_LANE]])
+// CHECK:   ret float [[VMULXS_F32_I]]
 float32_t test_vmulxs_lane_f32(float32_t a, float32x2_t b) {
   return vmulxs_lane_f32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vmulxs_laneq_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <4 x float> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[VMULXS_F32_I:%.*]] = call float @llvm.aarch64.neon.fmulx.f32(float [[A:%.*]], float [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret float [[VMULXS_F32_I]]
-//
+// CHECK-LABEL: define{{.*}} float @test_vmulxs_laneq_f32(float %a, <4 x float> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <4 x float> %b, i32 3
+// CHECK:   [[VMULXS_F32_I:%.*]] = call float @llvm.aarch64.neon.fmulx.f32(float %a, float [[VGETQ_LANE]])
+// CHECK:   ret float [[VMULXS_F32_I]]
 float32_t test_vmulxs_laneq_f32(float32_t a, float32x4_t b) {
   return vmulxs_laneq_f32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vmulxd_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[B:%.*]], i32 0
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[A:%.*]], double [[VGET_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret double [[VMULXD_F64_I]]
-//
+// CHECK-LABEL: define{{.*}} double @test_vmulxd_lane_f64(double %a, <1 x double> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> %b, i32 0
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double %a, double [[VGET_LANE]])
+// CHECK:   ret double [[VMULXD_F64_I]]
 float64_t test_vmulxd_lane_f64(float64_t a, float64x1_t b) {
   return vmulxd_lane_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vmulxd_laneq_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[A:%.*]], double [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret double [[VMULXD_F64_I]]
-//
+// CHECK-LABEL: define{{.*}} double @test_vmulxd_laneq_f64(double %a, <2 x double> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x double> %b, i32 1
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double %a, double [[VGETQ_LANE]])
+// CHECK:   ret double [[VMULXD_F64_I]]
 float64_t test_vmulxd_laneq_f64(float64_t a, float64x2_t b) {
   return vmulxd_laneq_f64(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vmulx_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[A:%.*]], i32 0
-// CHECK-NEXT:    [[VGET_LANE3:%.*]] = extractelement <1 x double> [[B:%.*]], i32 0
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGET_LANE3]]) [[ATTR5]]
-// CHECK-NEXT:    [[VSET_LANE:%.*]] = insertelement <1 x double> [[A]], double [[VMULXD_F64_I]], i32 0
-// CHECK-NEXT:    ret <1 x double> [[VSET_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vmulx_lane_f64(<1 x double> %a, <1 x double> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> %a, i32 0
+// CHECK:   [[VGET_LANE6:%.*]] = extractelement <1 x double> %b, i32 0
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGET_LANE6]])
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> %a, double [[VMULXD_F64_I]], i32 0
+// CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_lane_f64(float64x1_t a, float64x1_t b) {
   return vmulx_lane_f64(a, b, 0);
 }
 
 
-// CHECK-LABEL: @test_vmulx_laneq_f64_0(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[A:%.*]], i32 0
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[B:%.*]], i32 0
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    [[VSET_LANE:%.*]] = insertelement <1 x double> [[A]], double [[VMULXD_F64_I]], i32 0
-// CHECK-NEXT:    ret <1 x double> [[VSET_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vmulx_laneq_f64_0(<1 x double> %a, <2 x double> %b) #1 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> %a, i32 0
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x double> %b, i32 0
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]])
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> %a, double [[VMULXD_F64_I]], i32 0
+// CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_laneq_f64_0(float64x1_t a, float64x2_t b) {
   return vmulx_laneq_f64(a, b, 0);
 }
 
-// CHECK-LABEL: @test_vmulx_laneq_f64_1(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[A:%.*]], i32 0
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    [[VSET_LANE:%.*]] = insertelement <1 x double> [[A]], double [[VMULXD_F64_I]], i32 0
-// CHECK-NEXT:    ret <1 x double> [[VSET_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vmulx_laneq_f64_1(<1 x double> %a, <2 x double> %b) #1 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> %a, i32 0
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x double> %b, i32 1
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]])
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> %a, double [[VMULXD_F64_I]], i32 0
+// CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_laneq_f64_1(float64x1_t a, float64x2_t b) {
   return vmulx_laneq_f64(a, b, 1);
 }
 
 
-// CHECK-LABEL: @test_vfmas_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x float> [[C:%.*]], i32 1
-// CHECK-NEXT:    [[TMP0:%.*]] = call float @llvm.fma.f32(float [[B:%.*]], float [[EXTRACT]], float [[A:%.*]])
-// CHECK-NEXT:    ret float [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} float @test_vfmas_lane_f32(float %a, float %b, <2 x float> %c) #0 {
+// CHECK:   [[EXTRACT:%.*]] = extractelement <2 x float> %c, i32 1
+// CHECK:   [[TMP2:%.*]] = call float @llvm.fma.f32(float %b, float [[EXTRACT]], float %a)
+// CHECK:   ret float [[TMP2]]
 float32_t test_vfmas_lane_f32(float32_t a, float32_t b, float32x2_t c) {
   return vfmas_lane_f32(a, b, c, 1);
 }
 
-// CHECK-LABEL: @test_vfmad_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <1 x double> [[C:%.*]], i32 0
-// CHECK-NEXT:    [[TMP0:%.*]] = call double @llvm.fma.f64(double [[B:%.*]], double [[EXTRACT]], double [[A:%.*]])
-// CHECK-NEXT:    ret double [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} double @test_vfmad_lane_f64(double %a, double %b, <1 x double> %c) #0 {
+// CHECK:   [[EXTRACT:%.*]] = extractelement <1 x double> %c, i32 0
+// CHECK:   [[TMP2:%.*]] = call double @llvm.fma.f64(double %b, double [[EXTRACT]], double %a)
+// CHECK:   ret double [[TMP2]]
 float64_t test_vfmad_lane_f64(float64_t a, float64_t b, float64x1_t c) {
   return vfmad_lane_f64(a, b, c, 0);
 }
 
-// CHECK-LABEL: @test_vfmad_laneq_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x double> [[C:%.*]], i32 1
-// CHECK-NEXT:    [[TMP0:%.*]] = call double @llvm.fma.f64(double [[B:%.*]], double [[EXTRACT]], double [[A:%.*]])
-// CHECK-NEXT:    ret double [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} double @test_vfmad_laneq_f64(double %a, double %b, <2 x double> %c) #1 {
+// CHECK:   [[EXTRACT:%.*]] = extractelement <2 x double> %c, i32 1
+// CHECK:   [[TMP2:%.*]] = call double @llvm.fma.f64(double %b, double [[EXTRACT]], double %a)
+// CHECK:   ret double [[TMP2]]
 float64_t test_vfmad_laneq_f64(float64_t a, float64_t b, float64x2_t c) {
   return vfmad_laneq_f64(a, b, c, 1);
 }
 
-// CHECK-LABEL: @test_vfmss_lane_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[FNEG:%.*]] = fneg float [[B:%.*]]
-// CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x float> [[C:%.*]], i32 1
-// CHECK-NEXT:    [[TMP0:%.*]] = call float @llvm.fma.f32(float [[FNEG]], float [[EXTRACT]], float [[A:%.*]])
-// CHECK-NEXT:    ret float [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} float @test_vfmss_lane_f32(float %a, float %b, <2 x float> %c) #0 {
+// CHECK:   [[SUB:%.*]] = fneg float %b
+// CHECK:   [[EXTRACT:%.*]] = extractelement <2 x float> %c, i32 1
+// CHECK:   [[TMP2:%.*]] = call float @llvm.fma.f32(float [[SUB]], float [[EXTRACT]], float %a)
+// CHECK:   ret float [[TMP2]]
 float32_t test_vfmss_lane_f32(float32_t a, float32_t b, float32x2_t c) {
   return vfmss_lane_f32(a, b, c, 1);
 }
 
-// CHECK-LABEL: @test_vfma_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x double>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
-// CHECK-NEXT:    [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
-// CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
-// CHECK-NEXT:    [[FMLA2:%.*]] = call <1 x double> @llvm.fma.v1f64(<1 x double> [[FMLA]], <1 x double> [[LANE]], <1 x double> [[FMLA1]])
-// CHECK-NEXT:    ret <1 x double> [[FMLA2]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vfma_lane_f64(<1 x double> %a, <1 x double> %b, <1 x double> %v) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x double> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <1 x double> %v to <8 x i8>
+// CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x double>
+// CHECK:   [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
+// CHECK:   [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
+// CHECK:   [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
+// CHECK:   [[FMLA2:%.*]] = call <1 x double> @llvm.fma.v1f64(<1 x double> [[FMLA]], <1 x double> [[LANE]], <1 x double> [[FMLA1]])
+// CHECK:   ret <1 x double> [[FMLA2]]
 float64x1_t test_vfma_lane_f64(float64x1_t a, float64x1_t b, float64x1_t v) {
   return vfma_lane_f64(a, b, v, 0);
 }
 
-// CHECK-LABEL: @test_vfms_lane_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[FNEG:%.*]] = fneg <1 x double> [[B:%.*]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[FNEG]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <1 x double> [[V:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x double>
-// CHECK-NEXT:    [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
-// CHECK-NEXT:    [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
-// CHECK-NEXT:    [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
-// CHECK-NEXT:    [[FMLA2:%.*]] = call <1 x double> @llvm.fma.v1f64(<1 x double> [[FMLA]], <1 x double> [[LANE]], <1 x double> [[FMLA1]])
-// CHECK-NEXT:    ret <1 x double> [[FMLA2]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vfms_lane_f64(<1 x double> %a, <1 x double> %b, <1 x double> %v) #0 {
+// CHECK:   [[SUB:%.*]] = fneg <1 x double> %b
+// CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x double> [[SUB]] to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <1 x double> %v to <8 x i8>
+// CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP2]] to <1 x double>
+// CHECK:   [[LANE:%.*]] = shufflevector <1 x double> [[TMP3]], <1 x double> [[TMP3]], <1 x i32> zeroinitializer
+// CHECK:   [[FMLA:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x double>
+// CHECK:   [[FMLA1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
+// CHECK:   [[FMLA2:%.*]] = call <1 x double> @llvm.fma.v1f64(<1 x double> [[FMLA]], <1 x double> [[LANE]], <1 x double> [[FMLA1]])
+// CHECK:   ret <1 x double> [[FMLA2]]
 float64x1_t test_vfms_lane_f64(float64x1_t a, float64x1_t b, float64x1_t v) {
   return vfms_lane_f64(a, b, v, 0);
 }
 
-// CHECK-LABEL: @test_vfma_laneq_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[B:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to double
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to double
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
-// CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x double> [[TMP5]], i32 0
-// CHECK-NEXT:    [[TMP6:%.*]] = call double @llvm.fma.f64(double [[TMP4]], double [[EXTRACT]], double [[TMP3]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast double [[TMP6]] to <1 x double>
-// CHECK-NEXT:    ret <1 x double> [[TMP7]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vfma_laneq_f64(<1 x double> %a, <1 x double> %b, <2 x double> %v) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x double> %b to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <2 x double> %v to <16 x i8>
+// CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to double
+// CHECK:   [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to double
+// CHECK:   [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
+// CHECK:   [[EXTRACT:%.*]] = extractelement <2 x double> [[TMP5]], i32 0
+// CHECK:   [[TMP6:%.*]] = call double @llvm.fma.f64(double [[TMP4]], double [[EXTRACT]], double [[TMP3]])
+// CHECK:   [[TMP7:%.*]] = bitcast double [[TMP6]] to <1 x double>
+// CHECK:   ret <1 x double> [[TMP7]]
 float64x1_t test_vfma_laneq_f64(float64x1_t a, float64x1_t b, float64x2_t v) {
   return vfma_laneq_f64(a, b, v, 0);
 }
 
-// CHECK-LABEL: @test_vfms_laneq_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[FNEG:%.*]] = fneg <1 x double> [[B:%.*]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x double> [[A:%.*]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x double> [[FNEG]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <2 x double> [[V:%.*]] to <16 x i8>
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to double
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to double
-// CHECK-NEXT:    [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
-// CHECK-NEXT:    [[EXTRACT:%.*]] = extractelement <2 x double> [[TMP5]], i32 0
-// CHECK-NEXT:    [[TMP6:%.*]] = call double @llvm.fma.f64(double [[TMP4]], double [[EXTRACT]], double [[TMP3]])
-// CHECK-NEXT:    [[TMP7:%.*]] = bitcast double [[TMP6]] to <1 x double>
-// CHECK-NEXT:    ret <1 x double> [[TMP7]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vfms_laneq_f64(<1 x double> %a, <1 x double> %b, <2 x double> %v) #1 {
+// CHECK:   [[SUB:%.*]] = fneg <1 x double> %b
+// CHECK:   [[TMP0:%.*]] = bitcast <1 x double> %a to <8 x i8>
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x double> [[SUB]] to <8 x i8>
+// CHECK:   [[TMP2:%.*]] = bitcast <2 x double> %v to <16 x i8>
+// CHECK:   [[TMP3:%.*]] = bitcast <8 x i8> [[TMP0]] to double
+// CHECK:   [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to double
+// CHECK:   [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to <2 x double>
+// CHECK:   [[EXTRACT:%.*]] = extractelement <2 x double> [[TMP5]], i32 0
+// CHECK:   [[TMP6:%.*]] = call double @llvm.fma.f64(double [[TMP4]], double [[EXTRACT]], double [[TMP3]])
+// CHECK:   [[TMP7:%.*]] = bitcast double [[TMP6]] to <1 x double>
+// CHECK:   ret <1 x double> [[TMP7]]
 float64x1_t test_vfms_laneq_f64(float64x1_t a, float64x1_t b, float64x2_t v) {
   return vfms_laneq_f64(a, b, v, 0);
 }
 
-// CHECK-LABEL: @test_vqdmullh_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[A:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[VGET_LANE]], i64 0
-// CHECK-NEXT:    [[VQDMULLH_S16_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]]) [[ATTR5]]
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i32> [[VQDMULLH_S16_I]], i64 0
-// CHECK-NEXT:    ret i32 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmullh_lane_s16(i16 %a, <4 x i16> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <4 x i16> %b, i32 3
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %a, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[VGET_LANE]], i64 0
+// CHECK:   [[VQDMULLH_S16_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[TMP4:%.*]] = extractelement <4 x i32> [[VQDMULLH_S16_I]], i64 0
+// CHECK:   ret i32 [[TMP4]]
 int32_t test_vqdmullh_lane_s16(int16_t a, int16x4_t b) {
   return vqdmullh_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vqdmulls_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <2 x i32> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[VQDMULLS_S32_I:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 [[A:%.*]], i32 [[VGET_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret i64 [[VQDMULLS_S32_I]]
-//
+// CHECK-LABEL: define{{.*}} i64 @test_vqdmulls_lane_s32(i32 %a, <2 x i32> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <2 x i32> %b, i32 1
+// CHECK:   [[VQDMULLS_S32_I:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 %a, i32 [[VGET_LANE]])
+// CHECK:   ret i64 [[VQDMULLS_S32_I]]
 int64_t test_vqdmulls_lane_s32(int32_t a, int32x2_t b) {
   return vqdmulls_lane_s32(a, b, 1);
 }
 
-// CHECK-LABEL: @test_vqdmullh_laneq_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[B:%.*]], i32 7
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[A:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[VGETQ_LANE]], i64 0
-// CHECK-NEXT:    [[VQDMULLH_S16_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]]) [[ATTR5]]
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i32> [[VQDMULLH_S16_I]], i64 0
-// CHECK-NEXT:    ret i32 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmullh_laneq_s16(i16 %a, <8 x i16> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <8 x i16> %b, i32 7
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %a, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[VGETQ_LANE]], i64 0
+// CHECK:   [[VQDMULLH_S16_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[TMP4:%.*]] = extractelement <4 x i32> [[VQDMULLH_S16_I]], i64 0
+// CHECK:   ret i32 [[TMP4]]
 int32_t test_vqdmullh_laneq_s16(int16_t a, int16x8_t b) {
   return vqdmullh_laneq_s16(a, b, 7);
 }
 
-// CHECK-LABEL: @test_vqdmulls_laneq_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <4 x i32> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[VQDMULLS_S32_I:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 [[A:%.*]], i32 [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret i64 [[VQDMULLS_S32_I]]
-//
+// CHECK-LABEL: define{{.*}} i64 @test_vqdmulls_laneq_s32(i32 %a, <4 x i32> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <4 x i32> %b, i32 3
+// CHECK:   [[VQDMULLS_S32_I:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 %a, i32 [[VGETQ_LANE]])
+// CHECK:   ret i64 [[VQDMULLS_S32_I]]
 int64_t test_vqdmulls_laneq_s32(int32_t a, int32x4_t b) {
   return vqdmulls_laneq_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vqdmulhh_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[A:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[VGET_LANE]], i64 0
-// CHECK-NEXT:    [[VQDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]]) [[ATTR5]]
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i16> [[VQDMULHH_S16_I]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i16 @test_vqdmulhh_lane_s16(i16 %a, <4 x i16> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <4 x i16> %b, i32 3
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %a, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[VGET_LANE]], i64 0
+// CHECK:   [[VQDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.v4i16(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[TMP4:%.*]] = extractelement <4 x i16> [[VQDMULHH_S16_I]], i64 0
+// CHECK:   ret i16 [[TMP4]]
 int16_t test_vqdmulhh_lane_s16(int16_t a, int16x4_t b) {
   return vqdmulhh_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vqdmulhs_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <2 x i32> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[VQDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqdmulh.i32(i32 [[A:%.*]], i32 [[VGET_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret i32 [[VQDMULHS_S32_I]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmulhs_lane_s32(i32 %a, <2 x i32> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <2 x i32> %b, i32 1
+// CHECK:   [[VQDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqdmulh.i32(i32 %a, i32 [[VGET_LANE]])
+// CHECK:   ret i32 [[VQDMULHS_S32_I]]
 int32_t test_vqdmulhs_lane_s32(int32_t a, int32x2_t b) {
   return vqdmulhs_lane_s32(a, b, 1);
 }
 
 
-// CHECK-LABEL: @test_vqdmulhh_laneq_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[B:%.*]], i32 7
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[A:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[VGETQ_LANE]], i64 0
-// CHECK-NEXT:    [[VQDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]]) [[ATTR5]]
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i16> [[VQDMULHH_S16_I]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i16 @test_vqdmulhh_laneq_s16(i16 %a, <8 x i16> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <8 x i16> %b, i32 7
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %a, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[VGETQ_LANE]], i64 0
+// CHECK:   [[VQDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqdmulh.v4i16(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[TMP4:%.*]] = extractelement <4 x i16> [[VQDMULHH_S16_I]], i64 0
+// CHECK:   ret i16 [[TMP4]]
 int16_t test_vqdmulhh_laneq_s16(int16_t a, int16x8_t b) {
   return vqdmulhh_laneq_s16(a, b, 7);
 }
 
 
-// CHECK-LABEL: @test_vqdmulhs_laneq_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <4 x i32> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[VQDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqdmulh.i32(i32 [[A:%.*]], i32 [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret i32 [[VQDMULHS_S32_I]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmulhs_laneq_s32(i32 %a, <4 x i32> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <4 x i32> %b, i32 3
+// CHECK:   [[VQDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqdmulh.i32(i32 %a, i32 [[VGETQ_LANE]])
+// CHECK:   ret i32 [[VQDMULHS_S32_I]]
 int32_t test_vqdmulhs_laneq_s32(int32_t a, int32x4_t b) {
   return vqdmulhs_laneq_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vqrdmulhh_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <4 x i16> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[A:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[VGET_LANE]], i64 0
-// CHECK-NEXT:    [[VQRDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]]) [[ATTR5]]
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i16> [[VQRDMULHH_S16_I]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i16 @test_vqrdmulhh_lane_s16(i16 %a, <4 x i16> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <4 x i16> %b, i32 3
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %a, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[VGET_LANE]], i64 0
+// CHECK:   [[VQRDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.v4i16(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[TMP4:%.*]] = extractelement <4 x i16> [[VQRDMULHH_S16_I]], i64 0
+// CHECK:   ret i16 [[TMP4]]
 int16_t test_vqrdmulhh_lane_s16(int16_t a, int16x4_t b) {
   return vqrdmulhh_lane_s16(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vqrdmulhs_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <2 x i32> [[B:%.*]], i32 1
-// CHECK-NEXT:    [[VQRDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqrdmulh.i32(i32 [[A:%.*]], i32 [[VGET_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret i32 [[VQRDMULHS_S32_I]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqrdmulhs_lane_s32(i32 %a, <2 x i32> %b) #0 {
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <2 x i32> %b, i32 1
+// CHECK:   [[VQRDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqrdmulh.i32(i32 %a, i32 [[VGET_LANE]])
+// CHECK:   ret i32 [[VQRDMULHS_S32_I]]
 int32_t test_vqrdmulhs_lane_s32(int32_t a, int32x2_t b) {
   return vqrdmulhs_lane_s32(a, b, 1);
 }
 
 
-// CHECK-LABEL: @test_vqrdmulhh_laneq_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <8 x i16> [[B:%.*]], i32 7
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[A:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[VGETQ_LANE]], i64 0
-// CHECK-NEXT:    [[VQRDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]]) [[ATTR5]]
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i16> [[VQRDMULHH_S16_I]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i16 @test_vqrdmulhh_laneq_s16(i16 %a, <8 x i16> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <8 x i16> %b, i32 7
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %a, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[VGETQ_LANE]], i64 0
+// CHECK:   [[VQRDMULHH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqrdmulh.v4i16(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[TMP4:%.*]] = extractelement <4 x i16> [[VQRDMULHH_S16_I]], i64 0
+// CHECK:   ret i16 [[TMP4]]
 int16_t test_vqrdmulhh_laneq_s16(int16_t a, int16x8_t b) {
   return vqrdmulhh_laneq_s16(a, b, 7);
 }
 
 
-// CHECK-LABEL: @test_vqrdmulhs_laneq_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <4 x i32> [[B:%.*]], i32 3
-// CHECK-NEXT:    [[VQRDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqrdmulh.i32(i32 [[A:%.*]], i32 [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    ret i32 [[VQRDMULHS_S32_I]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqrdmulhs_laneq_s32(i32 %a, <4 x i32> %b) #1 {
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <4 x i32> %b, i32 3
+// CHECK:   [[VQRDMULHS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqrdmulh.i32(i32 %a, i32 [[VGETQ_LANE]])
+// CHECK:   ret i32 [[VQRDMULHS_S32_I]]
 int32_t test_vqrdmulhs_laneq_s32(int32_t a, int32x4_t b) {
   return vqrdmulhs_laneq_s32(a, b, 3);
 }
 
-// CHECK-LABEL: @test_vqdmlalh_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <4 x i16> [[C:%.*]], i32 3
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
-// CHECK-NEXT:    [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqadd.i32(i32 [[A:%.*]], i32 [[LANE0]])
-// CHECK-NEXT:    ret i32 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmlalh_lane_s16(i32 %a, i16 %b, <4 x i16> %c) #0 {
+// CHECK:   [[LANE:%.*]] = extractelement <4 x i16> %c, i32 3
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %b, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
+// CHECK:   [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
+// CHECK:   [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqadd.i32(i32 %a, i32 [[LANE0]])
+// CHECK:   ret i32 [[VQDMLXL1]]
 int32_t test_vqdmlalh_lane_s16(int32_t a, int16_t b, int16x4_t c) {
   return vqdmlalh_lane_s16(a, b, c, 3);
 }
 
-// CHECK-LABEL: @test_vqdmlals_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <2 x i32> [[C:%.*]], i32 1
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 [[B:%.*]], i32 [[LANE]])
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqadd.i64(i64 [[A:%.*]], i64 [[VQDMLXL]])
-// CHECK-NEXT:    ret i64 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i64 @test_vqdmlals_lane_s32(i64 %a, i32 %b, <2 x i32> %c) #0 {
+// CHECK:   [[LANE:%.*]] = extractelement <2 x i32> %c, i32 1
+// CHECK:   [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 %b, i32 [[LANE]])
+// CHECK:   [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqadd.i64(i64 %a, i64 [[VQDMLXL]])
+// CHECK:   ret i64 [[VQDMLXL1]]
 int64_t test_vqdmlals_lane_s32(int64_t a, int32_t b, int32x2_t c) {
   return vqdmlals_lane_s32(a, b, c, 1);
 }
 
-// CHECK-LABEL: @test_vqdmlalh_laneq_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <8 x i16> [[C:%.*]], i32 7
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
-// CHECK-NEXT:    [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqadd.i32(i32 [[A:%.*]], i32 [[LANE0]])
-// CHECK-NEXT:    ret i32 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmlalh_laneq_s16(i32 %a, i16 %b, <8 x i16> %c) #1 {
+// CHECK:   [[LANE:%.*]] = extractelement <8 x i16> %c, i32 7
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %b, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
+// CHECK:   [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
+// CHECK:   [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqadd.i32(i32 %a, i32 [[LANE0]])
+// CHECK:   ret i32 [[VQDMLXL1]]
 int32_t test_vqdmlalh_laneq_s16(int32_t a, int16_t b, int16x8_t c) {
   return vqdmlalh_laneq_s16(a, b, c, 7);
 }
 
-// CHECK-LABEL: @test_vqdmlals_laneq_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <4 x i32> [[C:%.*]], i32 3
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 [[B:%.*]], i32 [[LANE]])
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqadd.i64(i64 [[A:%.*]], i64 [[VQDMLXL]])
-// CHECK-NEXT:    ret i64 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i64 @test_vqdmlals_laneq_s32(i64 %a, i32 %b, <4 x i32> %c) #1 {
+// CHECK:   [[LANE:%.*]] = extractelement <4 x i32> %c, i32 3
+// CHECK:   [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 %b, i32 [[LANE]])
+// CHECK:   [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqadd.i64(i64 %a, i64 [[VQDMLXL]])
+// CHECK:   ret i64 [[VQDMLXL1]]
 int64_t test_vqdmlals_laneq_s32(int64_t a, int32_t b, int32x4_t c) {
   return vqdmlals_laneq_s32(a, b, c, 3);
 }
 
-// CHECK-LABEL: @test_vqdmlslh_lane_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <4 x i16> [[C:%.*]], i32 3
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
-// CHECK-NEXT:    [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqsub.i32(i32 [[A:%.*]], i32 [[LANE0]])
-// CHECK-NEXT:    ret i32 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmlslh_lane_s16(i32 %a, i16 %b, <4 x i16> %c) #0 {
+// CHECK:   [[LANE:%.*]] = extractelement <4 x i16> %c, i32 3
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %b, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
+// CHECK:   [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
+// CHECK:   [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqsub.i32(i32 %a, i32 [[LANE0]])
+// CHECK:   ret i32 [[VQDMLXL1]]
 int32_t test_vqdmlslh_lane_s16(int32_t a, int16_t b, int16x4_t c) {
   return vqdmlslh_lane_s16(a, b, c, 3);
 }
 
-// CHECK-LABEL: @test_vqdmlsls_lane_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <2 x i32> [[C:%.*]], i32 1
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 [[B:%.*]], i32 [[LANE]])
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqsub.i64(i64 [[A:%.*]], i64 [[VQDMLXL]])
-// CHECK-NEXT:    ret i64 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i64 @test_vqdmlsls_lane_s32(i64 %a, i32 %b, <2 x i32> %c) #0 {
+// CHECK:   [[LANE:%.*]] = extractelement <2 x i32> %c, i32 1
+// CHECK:   [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 %b, i32 [[LANE]])
+// CHECK:   [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqsub.i64(i64 %a, i64 [[VQDMLXL]])
+// CHECK:   ret i64 [[VQDMLXL1]]
 int64_t test_vqdmlsls_lane_s32(int64_t a, int32_t b, int32x2_t c) {
   return vqdmlsls_lane_s32(a, b, c, 1);
 }
 
-// CHECK-LABEL: @test_vqdmlslh_laneq_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <8 x i16> [[C:%.*]], i32 7
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> undef, i16 [[B:%.*]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
-// CHECK-NEXT:    [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqsub.i32(i32 [[A:%.*]], i32 [[LANE0]])
-// CHECK-NEXT:    ret i32 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i32 @test_vqdmlslh_laneq_s16(i32 %a, i16 %b, <8 x i16> %c) #1 {
+// CHECK:   [[LANE:%.*]] = extractelement <8 x i16> %c, i32 7
+// CHECK:   [[TMP2:%.*]] = insertelement <4 x i16> undef, i16 %b, i64 0
+// CHECK:   [[TMP3:%.*]] = insertelement <4 x i16> undef, i16 [[LANE]], i64 0
+// CHECK:   [[VQDMLXL:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> [[TMP2]], <4 x i16> [[TMP3]])
+// CHECK:   [[LANE0:%.*]] = extractelement <4 x i32> [[VQDMLXL]], i64 0
+// CHECK:   [[VQDMLXL1:%.*]] = call i32 @llvm.aarch64.neon.sqsub.i32(i32 %a, i32 [[LANE0]])
+// CHECK:   ret i32 [[VQDMLXL1]]
 int32_t test_vqdmlslh_laneq_s16(int32_t a, int16_t b, int16x8_t c) {
   return vqdmlslh_laneq_s16(a, b, c, 7);
 }
 
-// CHECK-LABEL: @test_vqdmlsls_laneq_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[LANE:%.*]] = extractelement <4 x i32> [[C:%.*]], i32 3
-// CHECK-NEXT:    [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 [[B:%.*]], i32 [[LANE]])
-// CHECK-NEXT:    [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqsub.i64(i64 [[A:%.*]], i64 [[VQDMLXL]])
-// CHECK-NEXT:    ret i64 [[VQDMLXL1]]
-//
+// CHECK-LABEL: define{{.*}} i64 @test_vqdmlsls_laneq_s32(i64 %a, i32 %b, <4 x i32> %c) #1 {
+// CHECK:   [[LANE:%.*]] = extractelement <4 x i32> %c, i32 3
+// CHECK:   [[VQDMLXL:%.*]] = call i64 @llvm.aarch64.neon.sqdmulls.scalar(i32 %b, i32 [[LANE]])
+// CHECK:   [[VQDMLXL1:%.*]] = call i64 @llvm.aarch64.neon.sqsub.i64(i64 %a, i64 [[VQDMLXL]])
+// CHECK:   ret i64 [[VQDMLXL1]]
 int64_t test_vqdmlsls_laneq_s32(int64_t a, int32_t b, int32x4_t c) {
   return vqdmlsls_laneq_s32(a, b, c, 3);
 }
 
-// CHECK-LABEL: @test_vmulx_lane_f64_0(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64 4599917171378402754 to <1 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64 4606655882138939123 to <1 x double>
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[TMP0]], i32 0
-// CHECK-NEXT:    [[VGET_LANE8:%.*]] = extractelement <1 x double> [[TMP1]], i32 0
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGET_LANE8]]) [[ATTR5]]
-// CHECK-NEXT:    [[VSET_LANE:%.*]] = insertelement <1 x double> [[TMP0]], double [[VMULXD_F64_I]], i32 0
-// CHECK-NEXT:    ret <1 x double> [[VSET_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vmulx_lane_f64_0() #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64 4599917171378402754 to <1 x double>
+// CHECK:   [[TMP1:%.*]] = bitcast i64 4606655882138939123 to <1 x double>
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> [[TMP0]], i32 0
+// CHECK:   [[VGET_LANE7:%.*]] = extractelement <1 x double> [[TMP1]], i32 0
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGET_LANE7]])
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> [[TMP0]], double [[VMULXD_F64_I]], i32 0
+// CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_lane_f64_0() {
       float64x1_t arg1;
       float64x1_t arg2;
@@ -508,17 +425,15 @@ float64x1_t test_vmulx_lane_f64_0() {
       return result;
 }
 
-// CHECK-LABEL: @test_vmulx_laneq_f64_2(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i64 4599917171378402754 to <1 x double>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i64 4606655882138939123 to <1 x double>
-// CHECK-NEXT:    [[SHUFFLE_I:%.*]] = shufflevector <1 x double> [[TMP0]], <1 x double> [[TMP1]], <2 x i32> <i32 0, i32 1>
-// CHECK-NEXT:    [[VGET_LANE:%.*]] = extractelement <1 x double> [[TMP0]], i32 0
-// CHECK-NEXT:    [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[SHUFFLE_I]], i32 1
-// CHECK-NEXT:    [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]]) [[ATTR5]]
-// CHECK-NEXT:    [[VSET_LANE:%.*]] = insertelement <1 x double> [[TMP0]], double [[VMULXD_F64_I]], i32 0
-// CHECK-NEXT:    ret <1 x double> [[VSET_LANE]]
-//
+// CHECK-LABEL: define{{.*}} <1 x double> @test_vmulx_laneq_f64_2() #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i64 4599917171378402754 to <1 x double>
+// CHECK:   [[TMP1:%.*]] = bitcast i64 4606655882138939123 to <1 x double>
+// CHECK:   [[SHUFFLE_I:%.*]] = shufflevector <1 x double> [[TMP0]], <1 x double> [[TMP1]], <2 x i32> <i32 0, i32 1>
+// CHECK:   [[VGET_LANE:%.*]] = extractelement <1 x double> [[TMP0]], i32 0
+// CHECK:   [[VGETQ_LANE:%.*]] = extractelement <2 x double> [[SHUFFLE_I]], i32 1
+// CHECK:   [[VMULXD_F64_I:%.*]] = call double @llvm.aarch64.neon.fmulx.f64(double [[VGET_LANE]], double [[VGETQ_LANE]])
+// CHECK:   [[VSET_LANE:%.*]] = insertelement <1 x double> [[TMP0]], double [[VMULXD_F64_I]], i32 0
+// CHECK:   ret <1 x double> [[VSET_LANE]]
 float64x1_t test_vmulx_laneq_f64_2() {
       float64x1_t arg1;
       float64x1_t arg2;
diff --git a/clang/test/CodeGen/aarch64-poly128.c b/clang/test/CodeGen/aarch64-poly128.c
index f55e36cf2f2f..113160f9d64b 100644
--- a/clang/test/CodeGen/aarch64-poly128.c
+++ b/clang/test/CodeGen/aarch64-poly128.c
@@ -1,4 +1,3 @@
-// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // REQUIRES: aarch64-registered-target
 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
 // RUN: -disable-O0-optnone -ffp-contract=fast -emit-llvm -o - %s | opt -S -mem2reg \
@@ -13,300 +12,238 @@
 
 #include <arm_neon.h>
 
-// CHECK-LABEL: @test_vstrq_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128* [[PTR:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i128*
-// CHECK-NEXT:    store i128 [[VAL:%.*]], i128* [[TMP1]], align 16
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_vstrq_p128(i128* %ptr, i128 %val) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128* %ptr to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i128*
+// CHECK:   store i128 %val, i128* [[TMP1]]
+// CHECK:   ret void
 void test_vstrq_p128(poly128_t * ptr, poly128_t val) {
   vstrq_p128(ptr, val);
 
 }
 
-// CHECK-LABEL: @test_vldrq_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128* [[PTR:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i128*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i128, i128* [[TMP1]], align 16
-// CHECK-NEXT:    ret i128 [[TMP2]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vldrq_p128(i128* %ptr) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128* %ptr to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i128*
+// CHECK:   [[TMP2:%.*]] = load i128, i128* [[TMP1]]
+// CHECK:   ret i128 [[TMP2]]
 poly128_t test_vldrq_p128(poly128_t * ptr) {
   return vldrq_p128(ptr);
 
 }
 
-// CHECK-LABEL: @test_ld_st_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128* [[PTR:%.*]] to i8*
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i128*
-// CHECK-NEXT:    [[TMP2:%.*]] = load i128, i128* [[TMP1]], align 16
-// CHECK-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i128, i128* [[PTR]], i64 1
-// CHECK-NEXT:    [[TMP3:%.*]] = bitcast i128* [[ADD_PTR]] to i8*
-// CHECK-NEXT:    [[TMP4:%.*]] = bitcast i8* [[TMP3]] to i128*
-// CHECK-NEXT:    store i128 [[TMP2]], i128* [[TMP4]], align 16
-// CHECK-NEXT:    ret void
-//
+// CHECK-LABEL: define{{.*}} void @test_ld_st_p128(i128* %ptr) #0 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128* %ptr to i8*
+// CHECK:   [[TMP1:%.*]] = bitcast i8* [[TMP0]] to i128*
+// CHECK:   [[TMP2:%.*]] = load i128, i128* [[TMP1]]
+// CHECK:   [[ADD_PTR:%.*]] = getelementptr inbounds i128, i128* %ptr, i64 1
+// CHECK:   [[TMP3:%.*]] = bitcast i128* [[ADD_PTR]] to i8*
+// CHECK:   [[TMP4:%.*]] = bitcast i8* [[TMP3]] to i128*
+// CHECK:   store i128 [[TMP2]], i128* [[TMP4]]
+// CHECK:   ret void
 void test_ld_st_p128(poly128_t * ptr) {
    vstrq_p128(ptr+1, vldrq_p128(ptr));
 
 }
 
-// CHECK-LABEL: @test_vmull_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[VMULL_P64_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.pmull64(i64 [[A:%.*]], i64 [[B:%.*]]) [[ATTR3:#.*]]
-// CHECK-NEXT:    [[VMULL_P641_I:%.*]] = bitcast <16 x i8> [[VMULL_P64_I]] to i128
-// CHECK-NEXT:    ret i128 [[VMULL_P641_I]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vmull_p64(i64 %a, i64 %b) #0 {
+// CHECK:   [[VMULL_P64_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.pmull64(i64 %a, i64 %b) #3
+// CHECK:   [[VMULL_P641_I:%.*]] = bitcast <16 x i8> [[VMULL_P64_I]] to i128
+// CHECK:   ret i128 [[VMULL_P641_I]]
 poly128_t test_vmull_p64(poly64_t a, poly64_t b) {
   return vmull_p64(a, b);
 }
 
-// CHECK-LABEL: @test_vmull_high_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[SHUFFLE_I_I:%.*]] = shufflevector <2 x i64> [[A:%.*]], <2 x i64> [[A]], <1 x i32> <i32 1>
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x i64> [[SHUFFLE_I_I]] to i64
-// CHECK-NEXT:    [[SHUFFLE_I7_I:%.*]] = shufflevector <2 x i64> [[B:%.*]], <2 x i64> [[B]], <1 x i32> <i32 1>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[SHUFFLE_I7_I]] to i64
-// CHECK-NEXT:    [[VMULL_P64_I_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.pmull64(i64 [[TMP0]], i64 [[TMP1]]) [[ATTR3]]
-// CHECK-NEXT:    [[VMULL_P641_I_I:%.*]] = bitcast <16 x i8> [[VMULL_P64_I_I]] to i128
-// CHECK-NEXT:    ret i128 [[VMULL_P641_I_I]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vmull_high_p64(<2 x i64> %a, <2 x i64> %b) #1 {
+// CHECK:   [[SHUFFLE_I_I:%.*]] = shufflevector <2 x i64> %a, <2 x i64> %a, <1 x i32> <i32 1>
+// CHECK:   [[TMP0:%.*]] = bitcast <1 x i64> [[SHUFFLE_I_I]] to i64
+// CHECK:   [[SHUFFLE_I7_I:%.*]] = shufflevector <2 x i64> %b, <2 x i64> %b, <1 x i32> <i32 1>
+// CHECK:   [[TMP1:%.*]] = bitcast <1 x i64> [[SHUFFLE_I7_I]] to i64
+// CHECK:   [[VMULL_P64_I_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.pmull64(i64 [[TMP0]], i64 [[TMP1]]) #3
+// CHECK:   [[VMULL_P641_I_I:%.*]] = bitcast <16 x i8> [[VMULL_P64_I_I]] to i128
+// CHECK:   ret i128 [[VMULL_P641_I_I]]
 poly128_t test_vmull_high_p64(poly64x2_t a, poly64x2_t b) {
   return vmull_high_p64(a, b);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_s8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <16 x i8> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_s8(<16 x i8> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <16 x i8> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_s8(int8x16_t a) {
   return vreinterpretq_p128_s8(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_s16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_s16(<8 x i16> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <8 x i16> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_s16(int16x8_t a) {
   return vreinterpretq_p128_s16(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_s32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_s32(<4 x i32> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <4 x i32> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_s32(int32x4_t a) {
   return vreinterpretq_p128_s32(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_s64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_s64(<2 x i64> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_s64(int64x2_t a) {
   return vreinterpretq_p128_s64(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_u8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <16 x i8> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_u8(<16 x i8> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <16 x i8> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_u8(uint8x16_t a) {
   return vreinterpretq_p128_u8(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_u16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_u16(<8 x i16> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <8 x i16> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_u16(uint16x8_t a) {
   return vreinterpretq_p128_u16(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_u32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_u32(<4 x i32> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <4 x i32> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_u32(uint32x4_t a) {
   return vreinterpretq_p128_u32(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_u64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_u64(<2 x i64> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_u64(uint64x2_t a) {
   return vreinterpretq_p128_u64(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_f32(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_f32(<4 x float> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <4 x float> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_f32(float32x4_t a) {
   return vreinterpretq_p128_f32(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_f64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x double> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_f64(<2 x double> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <2 x double> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_f64(float64x2_t a) {
   return vreinterpretq_p128_f64(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_p8(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <16 x i8> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_p8(<16 x i8> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <16 x i8> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_p8(poly8x16_t a) {
   return vreinterpretq_p128_p8(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_p16(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_p16(<8 x i16> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <8 x i16> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_p16(poly16x8_t a) {
   return vreinterpretq_p128_p16(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p128_p64(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A:%.*]] to i128
-// CHECK-NEXT:    ret i128 [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} i128 @test_vreinterpretq_p128_p64(<2 x i64> %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast <2 x i64> %a to i128
+// CHECK:   ret i128 [[TMP0]]
 poly128_t test_vreinterpretq_p128_p64(poly64x2_t a) {
   return vreinterpretq_p128_p64(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_s8_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    ret <16 x i8> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vreinterpretq_s8_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <16 x i8>
+// CHECK:   ret <16 x i8> [[TMP0]]
 int8x16_t test_vreinterpretq_s8_p128(poly128_t a) {
   return vreinterpretq_s8_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_s16_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    ret <8 x i16> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vreinterpretq_s16_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <8 x i16>
+// CHECK:   ret <8 x i16> [[TMP0]]
 int16x8_t test_vreinterpretq_s16_p128(poly128_t  a) {
   return vreinterpretq_s16_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_s32_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <4 x i32>
-// CHECK-NEXT:    ret <4 x i32> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i32> @test_vreinterpretq_s32_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <4 x i32>
+// CHECK:   ret <4 x i32> [[TMP0]]
 int32x4_t test_vreinterpretq_s32_p128(poly128_t a) {
   return vreinterpretq_s32_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_s64_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <2 x i64>
-// CHECK-NEXT:    ret <2 x i64> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vreinterpretq_s64_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <2 x i64>
+// CHECK:   ret <2 x i64> [[TMP0]]
 int64x2_t test_vreinterpretq_s64_p128(poly128_t  a) {
   return vreinterpretq_s64_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_u8_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    ret <16 x i8> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vreinterpretq_u8_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <16 x i8>
+// CHECK:   ret <16 x i8> [[TMP0]]
 uint8x16_t test_vreinterpretq_u8_p128(poly128_t  a) {
   return vreinterpretq_u8_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_u16_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    ret <8 x i16> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vreinterpretq_u16_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <8 x i16>
+// CHECK:   ret <8 x i16> [[TMP0]]
 uint16x8_t test_vreinterpretq_u16_p128(poly128_t  a) {
   return vreinterpretq_u16_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_u32_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <4 x i32>
-// CHECK-NEXT:    ret <4 x i32> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <4 x i32> @test_vreinterpretq_u32_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <4 x i32>
+// CHECK:   ret <4 x i32> [[TMP0]]
 uint32x4_t test_vreinterpretq_u32_p128(poly128_t  a) {
   return vreinterpretq_u32_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_u64_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <2 x i64>
-// CHECK-NEXT:    ret <2 x i64> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vreinterpretq_u64_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <2 x i64>
+// CHECK:   ret <2 x i64> [[TMP0]]
 uint64x2_t test_vreinterpretq_u64_p128(poly128_t  a) {
   return vreinterpretq_u64_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_f32_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <4 x float>
-// CHECK-NEXT:    ret <4 x float> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <4 x float> @test_vreinterpretq_f32_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <4 x float>
+// CHECK:   ret <4 x float> [[TMP0]]
 float32x4_t test_vreinterpretq_f32_p128(poly128_t  a) {
   return vreinterpretq_f32_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_f64_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <2 x double>
-// CHECK-NEXT:    ret <2 x double> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <2 x double> @test_vreinterpretq_f64_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <2 x double>
+// CHECK:   ret <2 x double> [[TMP0]]
 float64x2_t test_vreinterpretq_f64_p128(poly128_t  a) {
   return vreinterpretq_f64_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p8_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <16 x i8>
-// CHECK-NEXT:    ret <16 x i8> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <16 x i8> @test_vreinterpretq_p8_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <16 x i8>
+// CHECK:   ret <16 x i8> [[TMP0]]
 poly8x16_t test_vreinterpretq_p8_p128(poly128_t  a) {
   return vreinterpretq_p8_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p16_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <8 x i16>
-// CHECK-NEXT:    ret <8 x i16> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <8 x i16> @test_vreinterpretq_p16_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <8 x i16>
+// CHECK:   ret <8 x i16> [[TMP0]]
 poly16x8_t test_vreinterpretq_p16_p128(poly128_t  a) {
   return vreinterpretq_p16_p128(a);
 }
 
-// CHECK-LABEL: @test_vreinterpretq_p64_p128(
-// CHECK-NEXT:  entry:
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast i128 [[A:%.*]] to <2 x i64>
-// CHECK-NEXT:    ret <2 x i64> [[TMP0]]
-//
+// CHECK-LABEL: define{{.*}} <2 x i64> @test_vreinterpretq_p64_p128(i128 %a) #1 {
+// CHECK:   [[TMP0:%.*]] = bitcast i128 %a to <2 x i64>
+// CHECK:   ret <2 x i64> [[TMP0]]
 poly64x2_t test_vreinterpretq_p64_p128(poly128_t  a) {
   return vreinterpretq_p64_p128(a);
 }
diff --git a/clang/test/CodeGen/aarch64-poly64.c b/clang/test/CodeGen/aarch64-poly64.c
index 1452e8aff6c2..ebc58b5840fa 100644
--- a/clang/test/CodeGen/aarch64-poly64.c
+++ b/clang/test/CodeGen/aarch64-poly64.c
@@ -600,4 +600,4 @@ poly64x2_t test_vsriq_n_p64(poly64x2_t a, poly64x2_t b) {
 
 // CHECK: attributes #0 ={{.*}}"min-legal-vector-width"="64"
 // CHECK: attributes #1 ={{.*}}"min-legal-vector-width"="128"
-// CHECK-NOT: attributes #2 ={{.*}}"min-legal-vector-width"="0"
+// CHECK: attributes #2 ={{.*}}"min-legal-vector-width"="0"
diff --git a/clang/test/CodeGenCXX/dllexport-ctor-closure.cpp b/clang/test/CodeGenCXX/dllexport-ctor-closure.cpp
index 882cc71e57f7..42401e5ecd5d 100644
--- a/clang/test/CodeGenCXX/dllexport-ctor-closure.cpp
+++ b/clang/test/CodeGenCXX/dllexport-ctor-closure.cpp
@@ -5,7 +5,7 @@
 
 struct CtorWithClosure {
   __declspec(dllexport) CtorWithClosure(...) {}
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FCtorWithClosure@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FCtorWithClosure@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 // CHECK:   %[[this_addr:.*]] = alloca %struct.CtorWithClosure*, align 4
 // CHECK:   store %struct.CtorWithClosure* %this, %struct.CtorWithClosure** %[[this_addr]], align 4
 // CHECK:   %[[this:.*]] = load %struct.CtorWithClosure*, %struct.CtorWithClosure** %[[this_addr]]
@@ -17,7 +17,7 @@ struct CtorWithClosureOutOfLine {
   __declspec(dllexport) CtorWithClosureOutOfLine(...);
 };
 CtorWithClosureOutOfLine::CtorWithClosureOutOfLine(...) {}
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FCtorWithClosureOutOfLine@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FCtorWithClosureOutOfLine@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 
 #define DELETE_IMPLICIT_MEMBERS(ClassName) \
     ClassName(ClassName &&) = delete; \
@@ -28,7 +28,7 @@ CtorWithClosureOutOfLine::CtorWithClosureOutOfLine(...) {}
 struct __declspec(dllexport) ClassWithClosure {
   DELETE_IMPLICIT_MEMBERS(ClassWithClosure);
   ClassWithClosure(...) {}
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FClassWithClosure@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FClassWithClosure@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 // CHECK:   %[[this_addr:.*]] = alloca %struct.ClassWithClosure*, align 4
 // CHECK:   store %struct.ClassWithClosure* %this, %struct.ClassWithClosure** %[[this_addr]], align 4
 // CHECK:   %[[this:.*]] = load %struct.ClassWithClosure*, %struct.ClassWithClosure** %[[this_addr]]
@@ -44,10 +44,10 @@ template struct __declspec(dllexport) TemplateWithClosure<char>;
 extern template struct TemplateWithClosure<int>;
 template struct __declspec(dllexport) TemplateWithClosure<int>;
 
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$TemplateWithClosure@D@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$TemplateWithClosure@D@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 // CHECK:   call {{.*}} @"??0?$TemplateWithClosure@D@@QAE@H@Z"({{.*}}, i32 1)
 
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$TemplateWithClosure@H@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$TemplateWithClosure@H@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 // CHECK:   call {{.*}} @"??0?$TemplateWithClosure@H@@QAE@H@Z"({{.*}}, i32 4)
 
 template <typename T> struct __declspec(dllexport) ExportedTemplateWithClosure {
@@ -55,7 +55,7 @@ template <typename T> struct __declspec(dllexport) ExportedTemplateWithClosure {
 };
 template <> ExportedTemplateWithClosure<int>::ExportedTemplateWithClosure(int); // Don't try to emit the closure for a declaration.
 template <> ExportedTemplateWithClosure<int>::ExportedTemplateWithClosure(int) {};
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$ExportedTemplateWithClosure@H@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$ExportedTemplateWithClosure@H@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 // CHECK:   call {{.*}} @"??0?$ExportedTemplateWithClosure@H@@QAE@H@Z"({{.*}}, i32 4)
 
 struct __declspec(dllexport) NestedOuter {
@@ -67,8 +67,8 @@ struct __declspec(dllexport) NestedOuter {
   };
 };
 
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FNestedOuter@@QAEXXZ"({{.*}}) comdat
-// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FNestedInner@NestedOuter@@QAEXXZ"({{.*}}) comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FNestedOuter@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
+// CHECK-LABEL: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FNestedInner@NestedOuter@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 
 struct HasDtor {
   ~HasDtor();
diff --git a/clang/test/CodeGenCXX/dllexport.cpp b/clang/test/CodeGenCXX/dllexport.cpp
index 6c84c21c5f35..a0f53c8e2e9f 100644
--- a/clang/test/CodeGenCXX/dllexport.cpp
+++ b/clang/test/CodeGenCXX/dllexport.cpp
@@ -535,7 +535,7 @@ struct SomeTemplate {
 // MSVC2013-DAG: define weak_odr dso_local dllexport {{.+}} @"??4?$SomeTemplate@H@@Q{{.+}}0@A{{.+}}0@@Z"
 struct __declspec(dllexport) InheritFromTemplate : SomeTemplate<int> {};
 
-// M32-DAG: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$SomeTemplate@H@@QAEXXZ"({{.*}}) comdat
+// M32-DAG: define weak_odr dso_local dllexport x86_thiscallcc void @"??_F?$SomeTemplate@H@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 
 namespace PR23801 {
 template <typename>
@@ -552,7 +552,7 @@ struct __declspec(dllexport) B {
 
 }
 //
-// M32-DAG: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FB@PR23801@@QAEXXZ"({{.*}}) comdat
+// M32-DAG: define weak_odr dso_local dllexport x86_thiscallcc void @"??_FB@PR23801@@QAEXXZ"({{.*}}) {{#[0-9]+}} comdat
 
 struct __declspec(dllexport) T {
   // Copy assignment operator: