[clang] 5a2e56b - [Clang][NeonEmitter] emit ret decl first for -Wdeclaration-after-statement
Nick Desaulniers via cfe-commits
cfe-commits at lists.llvm.org
Wed Mar 23 09:41:02 PDT 2022
Author: Nick Desaulniers
Date: 2022-03-23T09:40:43-07:00
New Revision: 5a2e56b70e2fa7ad0d82e54bc4c741b16f05e475
URL: https://github.com/llvm/llvm-project/commit/5a2e56b70e2fa7ad0d82e54bc4c741b16f05e475
DIFF: https://github.com/llvm/llvm-project/commit/5a2e56b70e2fa7ad0d82e54bc4c741b16f05e475.diff
LOG: [Clang][NeonEmitter] emit ret decl first for -Wdeclaration-after-statement
The generated arm_neon.h header isn't -Wdeclaration-after-statement
compliant when targeting -mbig-endian. Update the generator to declare
the return value, if any, first before any other arguments that might
need to be "reversed" from little endian to big.
Another approach would have been to try to ignore this warning in system
headers, though that might not be precise for tokens involved in macro
expansion. See also: https://reviews.llvm.org/D116833#3236209.
Link: https://github.com/ClangBuiltLinux/linux/issues/1603
Fixes: https://github.com/llvm/llvm-project/issues/54062
Reviewed By: DavidSpickett
Differential Revision: https://reviews.llvm.org/D122189
Added:
clang/test/Sema/arm-neon-decl-after-stmt.c
Modified:
clang/test/CodeGen/aarch64-neon-ldst-one.c
clang/test/CodeGen/arm_neon_intrinsics.c
clang/utils/TableGen/NeonEmitter.cpp
Removed:
################################################################################
diff --git a/clang/test/CodeGen/aarch64-neon-ldst-one.c b/clang/test/CodeGen/aarch64-neon-ldst-one.c
index 1ecdd3e73724e..12f364b4425d0 100644
--- a/clang/test/CodeGen/aarch64-neon-ldst-one.c
+++ b/clang/test/CodeGen/aarch64-neon-ldst-one.c
@@ -947,8 +947,8 @@ poly64x1_t test_vld1_lane_p64(poly64_t *a, poly64x1_t b) {
// CHECK-LABEL: define{{.*}} %struct.int8x16x2_t @test_vld2q_lane_s8(i8* noundef %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: [[__S1:%.*]] = 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*
@@ -976,8 +976,8 @@ int8x16x2_t test_vld2q_lane_s8(int8_t const * ptr, int8x16x2_t src) {
// CHECK-LABEL: define{{.*}} %struct.uint8x16x2_t @test_vld2q_lane_u8(i8* noundef %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: [[__S1:%.*]] = 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*
@@ -1005,8 +1005,8 @@ uint8x16x2_t test_vld2q_lane_u8(uint8_t const * ptr, uint8x16x2_t src) {
// CHECK-LABEL: define{{.*}} %struct.poly8x16x2_t @test_vld2q_lane_p8(i8* noundef %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: [[__S1:%.*]] = 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*
@@ -1034,8 +1034,8 @@ poly8x16x2_t test_vld2q_lane_p8(poly8_t const * ptr, poly8x16x2_t src) {
// CHECK-LABEL: define{{.*}} %struct.int8x16x3_t @test_vld3q_lane_s8(i8* noundef %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: [[__S1:%.*]] = 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*
@@ -1066,8 +1066,8 @@ int8x16x3_t test_vld3q_lane_s8(int8_t const * ptr, int8x16x3_t src) {
// CHECK-LABEL: define{{.*}} %struct.uint8x16x3_t @test_vld3q_lane_u8(i8* noundef %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: [[__S1:%.*]] = 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*
@@ -1098,8 +1098,8 @@ uint8x16x3_t test_vld3q_lane_u8(uint8_t const * ptr, uint8x16x3_t src) {
// CHECK-LABEL: define{{.*}} %struct.uint16x8x2_t @test_vld2q_lane_u16(i16* noundef %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: [[__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*
@@ -1132,8 +1132,8 @@ uint16x8x2_t test_vld2q_lane_u16(uint16_t *a, uint16x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint32x4x2_t @test_vld2q_lane_u32(i32* noundef %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: [[__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*
@@ -1166,8 +1166,8 @@ uint32x4x2_t test_vld2q_lane_u32(uint32_t *a, uint32x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint64x2x2_t @test_vld2q_lane_u64(i64* noundef %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: [[__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*
@@ -1200,8 +1200,8 @@ uint64x2x2_t test_vld2q_lane_u64(uint64_t *a, uint64x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int16x8x2_t @test_vld2q_lane_s16(i16* noundef %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: [[__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*
@@ -1234,8 +1234,8 @@ int16x8x2_t test_vld2q_lane_s16(int16_t *a, int16x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int32x4x2_t @test_vld2q_lane_s32(i32* noundef %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: [[__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*
@@ -1268,8 +1268,8 @@ int32x4x2_t test_vld2q_lane_s32(int32_t *a, int32x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int64x2x2_t @test_vld2q_lane_s64(i64* noundef %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: [[__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*
@@ -1302,8 +1302,8 @@ int64x2x2_t test_vld2q_lane_s64(int64_t *a, int64x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.float16x8x2_t @test_vld2q_lane_f16(half* noundef %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: [[__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*
@@ -1336,8 +1336,8 @@ float16x8x2_t test_vld2q_lane_f16(float16_t *a, float16x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.float32x4x2_t @test_vld2q_lane_f32(float* noundef %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: [[__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*
@@ -1370,8 +1370,8 @@ float32x4x2_t test_vld2q_lane_f32(float32_t *a, float32x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.float64x2x2_t @test_vld2q_lane_f64(double* noundef %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: [[__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*
@@ -1404,8 +1404,8 @@ float64x2x2_t test_vld2q_lane_f64(float64_t *a, float64x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly16x8x2_t @test_vld2q_lane_p16(i16* noundef %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: [[__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*
@@ -1438,8 +1438,8 @@ poly16x8x2_t test_vld2q_lane_p16(poly16_t *a, poly16x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly64x2x2_t @test_vld2q_lane_p64(i64* noundef %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: [[__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*
@@ -1472,8 +1472,8 @@ poly64x2x2_t test_vld2q_lane_p64(poly64_t *a, poly64x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint8x8x2_t @test_vld2_lane_u8(i8* noundef %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: [[__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*
@@ -1501,8 +1501,8 @@ uint8x8x2_t test_vld2_lane_u8(uint8_t *a, uint8x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint16x4x2_t @test_vld2_lane_u16(i16* noundef %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: [[__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*
@@ -1535,8 +1535,8 @@ uint16x4x2_t test_vld2_lane_u16(uint16_t *a, uint16x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint32x2x2_t @test_vld2_lane_u32(i32* noundef %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: [[__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*
@@ -1569,8 +1569,8 @@ uint32x2x2_t test_vld2_lane_u32(uint32_t *a, uint32x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint64x1x2_t @test_vld2_lane_u64(i64* noundef %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: [[__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*
@@ -1603,8 +1603,8 @@ uint64x1x2_t test_vld2_lane_u64(uint64_t *a, uint64x1x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int8x8x2_t @test_vld2_lane_s8(i8* noundef %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: [[__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*
@@ -1632,8 +1632,8 @@ int8x8x2_t test_vld2_lane_s8(int8_t *a, int8x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int16x4x2_t @test_vld2_lane_s16(i16* noundef %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: [[__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*
@@ -1666,8 +1666,8 @@ int16x4x2_t test_vld2_lane_s16(int16_t *a, int16x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int32x2x2_t @test_vld2_lane_s32(i32* noundef %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: [[__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*
@@ -1700,8 +1700,8 @@ int32x2x2_t test_vld2_lane_s32(int32_t *a, int32x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.int64x1x2_t @test_vld2_lane_s64(i64* noundef %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: [[__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*
@@ -1734,8 +1734,8 @@ int64x1x2_t test_vld2_lane_s64(int64_t *a, int64x1x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.float16x4x2_t @test_vld2_lane_f16(half* noundef %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: [[__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*
@@ -1768,8 +1768,8 @@ float16x4x2_t test_vld2_lane_f16(float16_t *a, float16x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.float32x2x2_t @test_vld2_lane_f32(float* noundef %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: [[__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*
@@ -1802,8 +1802,8 @@ float32x2x2_t test_vld2_lane_f32(float32_t *a, float32x2x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.float64x1x2_t @test_vld2_lane_f64(double* noundef %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: [[__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*
@@ -1836,8 +1836,8 @@ float64x1x2_t test_vld2_lane_f64(float64_t *a, float64x1x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly8x8x2_t @test_vld2_lane_p8(i8* noundef %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: [[__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*
@@ -1865,8 +1865,8 @@ poly8x8x2_t test_vld2_lane_p8(poly8_t *a, poly8x8x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly16x4x2_t @test_vld2_lane_p16(i16* noundef %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: [[__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*
@@ -1899,8 +1899,8 @@ poly16x4x2_t test_vld2_lane_p16(poly16_t *a, poly16x4x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly64x1x2_t @test_vld2_lane_p64(i64* noundef %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: [[__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*
@@ -1933,8 +1933,8 @@ poly64x1x2_t test_vld2_lane_p64(poly64_t *a, poly64x1x2_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint16x8x3_t @test_vld3q_lane_u16(i16* noundef %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: [[__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*
@@ -1972,8 +1972,8 @@ uint16x8x3_t test_vld3q_lane_u16(uint16_t *a, uint16x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint32x4x3_t @test_vld3q_lane_u32(i32* noundef %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: [[__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*
@@ -2011,8 +2011,8 @@ uint32x4x3_t test_vld3q_lane_u32(uint32_t *a, uint32x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint64x2x3_t @test_vld3q_lane_u64(i64* noundef %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: [[__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*
@@ -2050,8 +2050,8 @@ uint64x2x3_t test_vld3q_lane_u64(uint64_t *a, uint64x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int16x8x3_t @test_vld3q_lane_s16(i16* noundef %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: [[__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*
@@ -2089,8 +2089,8 @@ int16x8x3_t test_vld3q_lane_s16(int16_t *a, int16x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int32x4x3_t @test_vld3q_lane_s32(i32* noundef %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: [[__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*
@@ -2128,8 +2128,8 @@ int32x4x3_t test_vld3q_lane_s32(int32_t *a, int32x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int64x2x3_t @test_vld3q_lane_s64(i64* noundef %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: [[__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*
@@ -2167,8 +2167,8 @@ int64x2x3_t test_vld3q_lane_s64(int64_t *a, int64x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.float16x8x3_t @test_vld3q_lane_f16(half* noundef %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: [[__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*
@@ -2206,8 +2206,8 @@ float16x8x3_t test_vld3q_lane_f16(float16_t *a, float16x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.float32x4x3_t @test_vld3q_lane_f32(float* noundef %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: [[__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*
@@ -2245,8 +2245,8 @@ float32x4x3_t test_vld3q_lane_f32(float32_t *a, float32x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.float64x2x3_t @test_vld3q_lane_f64(double* noundef %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: [[__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*
@@ -2284,8 +2284,8 @@ float64x2x3_t test_vld3q_lane_f64(float64_t *a, float64x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly8x16x3_t @test_vld3q_lane_p8(i8* noundef %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: [[__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*
@@ -2316,8 +2316,8 @@ poly8x16x3_t test_vld3q_lane_p8(poly8_t *a, poly8x16x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly16x8x3_t @test_vld3q_lane_p16(i16* noundef %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: [[__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*
@@ -2355,8 +2355,8 @@ poly16x8x3_t test_vld3q_lane_p16(poly16_t *a, poly16x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly64x2x3_t @test_vld3q_lane_p64(i64* noundef %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: [[__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*
@@ -2394,8 +2394,8 @@ poly64x2x3_t test_vld3q_lane_p64(poly64_t *a, poly64x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint8x8x3_t @test_vld3_lane_u8(i8* noundef %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: [[__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*
@@ -2426,8 +2426,8 @@ uint8x8x3_t test_vld3_lane_u8(uint8_t *a, uint8x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint16x4x3_t @test_vld3_lane_u16(i16* noundef %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: [[__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*
@@ -2465,8 +2465,8 @@ uint16x4x3_t test_vld3_lane_u16(uint16_t *a, uint16x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint32x2x3_t @test_vld3_lane_u32(i32* noundef %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: [[__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*
@@ -2504,8 +2504,8 @@ uint32x2x3_t test_vld3_lane_u32(uint32_t *a, uint32x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint64x1x3_t @test_vld3_lane_u64(i64* noundef %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: [[__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*
@@ -2543,8 +2543,8 @@ uint64x1x3_t test_vld3_lane_u64(uint64_t *a, uint64x1x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int8x8x3_t @test_vld3_lane_s8(i8* noundef %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: [[__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*
@@ -2575,8 +2575,8 @@ int8x8x3_t test_vld3_lane_s8(int8_t *a, int8x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int16x4x3_t @test_vld3_lane_s16(i16* noundef %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: [[__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*
@@ -2614,8 +2614,8 @@ int16x4x3_t test_vld3_lane_s16(int16_t *a, int16x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int32x2x3_t @test_vld3_lane_s32(i32* noundef %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: [[__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*
@@ -2653,8 +2653,8 @@ int32x2x3_t test_vld3_lane_s32(int32_t *a, int32x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.int64x1x3_t @test_vld3_lane_s64(i64* noundef %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: [[__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*
@@ -2692,8 +2692,8 @@ int64x1x3_t test_vld3_lane_s64(int64_t *a, int64x1x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.float16x4x3_t @test_vld3_lane_f16(half* noundef %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: [[__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*
@@ -2731,8 +2731,8 @@ float16x4x3_t test_vld3_lane_f16(float16_t *a, float16x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.float32x2x3_t @test_vld3_lane_f32(float* noundef %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: [[__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*
@@ -2770,8 +2770,8 @@ float32x2x3_t test_vld3_lane_f32(float32_t *a, float32x2x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.float64x1x3_t @test_vld3_lane_f64(double* noundef %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: [[__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*
@@ -2809,8 +2809,8 @@ float64x1x3_t test_vld3_lane_f64(float64_t *a, float64x1x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly8x8x3_t @test_vld3_lane_p8(i8* noundef %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: [[__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*
@@ -2841,8 +2841,8 @@ poly8x8x3_t test_vld3_lane_p8(poly8_t *a, poly8x8x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly16x4x3_t @test_vld3_lane_p16(i16* noundef %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: [[__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*
@@ -2880,8 +2880,8 @@ poly16x4x3_t test_vld3_lane_p16(poly16_t *a, poly16x4x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly64x1x3_t @test_vld3_lane_p64(i64* noundef %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: [[__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*
@@ -2919,8 +2919,8 @@ poly64x1x3_t test_vld3_lane_p64(poly64_t *a, poly64x1x3_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint8x16x4_t @test_vld4q_lane_u8(i8* noundef %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: [[__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*
@@ -2954,8 +2954,8 @@ uint8x16x4_t test_vld4q_lane_u8(uint8_t *a, uint8x16x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint16x8x4_t @test_vld4q_lane_u16(i16* noundef %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: [[__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*
@@ -2998,8 +2998,8 @@ uint16x8x4_t test_vld4q_lane_u16(uint16_t *a, uint16x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint32x4x4_t @test_vld4q_lane_u32(i32* noundef %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: [[__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*
@@ -3042,8 +3042,8 @@ uint32x4x4_t test_vld4q_lane_u32(uint32_t *a, uint32x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint64x2x4_t @test_vld4q_lane_u64(i64* noundef %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: [[__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*
@@ -3086,8 +3086,8 @@ uint64x2x4_t test_vld4q_lane_u64(uint64_t *a, uint64x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int8x16x4_t @test_vld4q_lane_s8(i8* noundef %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: [[__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*
@@ -3121,8 +3121,8 @@ int8x16x4_t test_vld4q_lane_s8(int8_t *a, int8x16x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int16x8x4_t @test_vld4q_lane_s16(i16* noundef %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: [[__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*
@@ -3165,8 +3165,8 @@ int16x8x4_t test_vld4q_lane_s16(int16_t *a, int16x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int32x4x4_t @test_vld4q_lane_s32(i32* noundef %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: [[__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*
@@ -3209,8 +3209,8 @@ int32x4x4_t test_vld4q_lane_s32(int32_t *a, int32x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int64x2x4_t @test_vld4q_lane_s64(i64* noundef %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: [[__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*
@@ -3253,8 +3253,8 @@ int64x2x4_t test_vld4q_lane_s64(int64_t *a, int64x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.float16x8x4_t @test_vld4q_lane_f16(half* noundef %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: [[__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*
@@ -3297,8 +3297,8 @@ float16x8x4_t test_vld4q_lane_f16(float16_t *a, float16x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.float32x4x4_t @test_vld4q_lane_f32(float* noundef %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: [[__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*
@@ -3341,8 +3341,8 @@ float32x4x4_t test_vld4q_lane_f32(float32_t *a, float32x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.float64x2x4_t @test_vld4q_lane_f64(double* noundef %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: [[__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*
@@ -3385,8 +3385,8 @@ float64x2x4_t test_vld4q_lane_f64(float64_t *a, float64x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly8x16x4_t @test_vld4q_lane_p8(i8* noundef %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: [[__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*
@@ -3420,8 +3420,8 @@ poly8x16x4_t test_vld4q_lane_p8(poly8_t *a, poly8x16x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly16x8x4_t @test_vld4q_lane_p16(i16* noundef %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: [[__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*
@@ -3464,8 +3464,8 @@ poly16x8x4_t test_vld4q_lane_p16(poly16_t *a, poly16x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly64x2x4_t @test_vld4q_lane_p64(i64* noundef %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: [[__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*
@@ -3508,8 +3508,8 @@ poly64x2x4_t test_vld4q_lane_p64(poly64_t *a, poly64x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint8x8x4_t @test_vld4_lane_u8(i8* noundef %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: [[__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*
@@ -3543,8 +3543,8 @@ uint8x8x4_t test_vld4_lane_u8(uint8_t *a, uint8x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint16x4x4_t @test_vld4_lane_u16(i16* noundef %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: [[__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*
@@ -3587,8 +3587,8 @@ uint16x4x4_t test_vld4_lane_u16(uint16_t *a, uint16x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint32x2x4_t @test_vld4_lane_u32(i32* noundef %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: [[__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*
@@ -3631,8 +3631,8 @@ uint32x2x4_t test_vld4_lane_u32(uint32_t *a, uint32x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.uint64x1x4_t @test_vld4_lane_u64(i64* noundef %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: [[__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*
@@ -3675,8 +3675,8 @@ uint64x1x4_t test_vld4_lane_u64(uint64_t *a, uint64x1x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int8x8x4_t @test_vld4_lane_s8(i8* noundef %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: [[__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*
@@ -3710,8 +3710,8 @@ int8x8x4_t test_vld4_lane_s8(int8_t *a, int8x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int16x4x4_t @test_vld4_lane_s16(i16* noundef %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: [[__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*
@@ -3754,8 +3754,8 @@ int16x4x4_t test_vld4_lane_s16(int16_t *a, int16x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int32x2x4_t @test_vld4_lane_s32(i32* noundef %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: [[__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*
@@ -3798,8 +3798,8 @@ int32x2x4_t test_vld4_lane_s32(int32_t *a, int32x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.int64x1x4_t @test_vld4_lane_s64(i64* noundef %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: [[__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*
@@ -3842,8 +3842,8 @@ int64x1x4_t test_vld4_lane_s64(int64_t *a, int64x1x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.float16x4x4_t @test_vld4_lane_f16(half* noundef %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: [[__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*
@@ -3886,8 +3886,8 @@ float16x4x4_t test_vld4_lane_f16(float16_t *a, float16x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.float32x2x4_t @test_vld4_lane_f32(float* noundef %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: [[__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*
@@ -3930,8 +3930,8 @@ float32x2x4_t test_vld4_lane_f32(float32_t *a, float32x2x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.float64x1x4_t @test_vld4_lane_f64(double* noundef %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: [[__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*
@@ -3974,8 +3974,8 @@ float64x1x4_t test_vld4_lane_f64(float64_t *a, float64x1x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly8x8x4_t @test_vld4_lane_p8(i8* noundef %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: [[__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*
@@ -4009,8 +4009,8 @@ poly8x8x4_t test_vld4_lane_p8(poly8_t *a, poly8x8x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly16x4x4_t @test_vld4_lane_p16(i16* noundef %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: [[__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*
@@ -4053,8 +4053,8 @@ poly16x4x4_t test_vld4_lane_p16(poly16_t *a, poly16x4x4_t b) {
// CHECK-LABEL: define{{.*}} %struct.poly64x1x4_t @test_vld4_lane_p64(i64* noundef %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: [[__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*
diff --git a/clang/test/CodeGen/arm_neon_intrinsics.c b/clang/test/CodeGen/arm_neon_intrinsics.c
index d7d22eb35a5ed..09ceae21a6ad8 100644
--- a/clang/test/CodeGen/arm_neon_intrinsics.c
+++ b/clang/test/CodeGen/arm_neon_intrinsics.c
@@ -4785,8 +4785,8 @@ poly16x4x2_t test_vld2_p16(poly16_t const * a) {
// CHECK-LABEL: @test_vld2q_lane_u16(
// 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: [[__S1:%.*]] = alloca %struct.uint16x8x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x2_t, %struct.uint16x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x i16>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4812,8 +4812,8 @@ uint16x8x2_t test_vld2q_lane_u16(uint16_t const * a, uint16x8x2_t b) {
// CHECK-LABEL: @test_vld2q_lane_u32(
// 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: [[__S1:%.*]] = alloca %struct.uint32x4x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x2_t, %struct.uint32x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x i32>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4839,8 +4839,8 @@ uint32x4x2_t test_vld2q_lane_u32(uint32_t const * a, uint32x4x2_t b) {
// CHECK-LABEL: @test_vld2q_lane_s16(
// 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: [[__S1:%.*]] = alloca %struct.int16x8x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x2_t, %struct.int16x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x i16>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4866,8 +4866,8 @@ int16x8x2_t test_vld2q_lane_s16(int16_t const * a, int16x8x2_t b) {
// CHECK-LABEL: @test_vld2q_lane_s32(
// 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: [[__S1:%.*]] = alloca %struct.int32x4x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x2_t, %struct.int32x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x i32>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4893,8 +4893,8 @@ int32x4x2_t test_vld2q_lane_s32(int32_t const * a, int32x4x2_t b) {
// CHECK-LABEL: @test_vld2q_lane_f16(
// 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: [[__S1:%.*]] = alloca %struct.float16x8x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x2_t, %struct.float16x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x half>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4920,8 +4920,8 @@ float16x8x2_t test_vld2q_lane_f16(float16_t const * a, float16x8x2_t b) {
// CHECK-LABEL: @test_vld2q_lane_f32(
// 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: [[__S1:%.*]] = alloca %struct.float32x4x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x2_t, %struct.float32x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x float>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4947,8 +4947,8 @@ float32x4x2_t test_vld2q_lane_f32(float32_t const * a, float32x4x2_t b) {
// CHECK-LABEL: @test_vld2q_lane_p16(
// 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: [[__S1:%.*]] = alloca %struct.poly16x8x2_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x2_t, %struct.poly16x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x i16>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 16
@@ -4974,8 +4974,8 @@ poly16x8x2_t test_vld2q_lane_p16(poly16_t const * a, poly16x8x2_t b) {
// CHECK-LABEL: @test_vld2_lane_u8(
// 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: [[__S1:%.*]] = alloca %struct.uint8x8x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x2_t, %struct.uint8x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x i8>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -4996,8 +4996,8 @@ uint8x8x2_t test_vld2_lane_u8(uint8_t const * a, uint8x8x2_t b) {
// CHECK-LABEL: @test_vld2_lane_u16(
// 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: [[__S1:%.*]] = alloca %struct.uint16x4x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x2_t, %struct.uint16x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x i16>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5023,8 +5023,8 @@ uint16x4x2_t test_vld2_lane_u16(uint16_t const * a, uint16x4x2_t b) {
// CHECK-LABEL: @test_vld2_lane_u32(
// 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: [[__S1:%.*]] = alloca %struct.uint32x2x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x2_t, %struct.uint32x2x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <2 x i32>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5050,8 +5050,8 @@ uint32x2x2_t test_vld2_lane_u32(uint32_t const * a, uint32x2x2_t b) {
// CHECK-LABEL: @test_vld2_lane_s8(
// 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: [[__S1:%.*]] = alloca %struct.int8x8x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x2_t, %struct.int8x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x i8>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5072,8 +5072,8 @@ int8x8x2_t test_vld2_lane_s8(int8_t const * a, int8x8x2_t b) {
// CHECK-LABEL: @test_vld2_lane_s16(
// 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: [[__S1:%.*]] = alloca %struct.int16x4x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x2_t, %struct.int16x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x i16>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5099,8 +5099,8 @@ int16x4x2_t test_vld2_lane_s16(int16_t const * a, int16x4x2_t b) {
// CHECK-LABEL: @test_vld2_lane_s32(
// 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: [[__S1:%.*]] = alloca %struct.int32x2x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x2_t, %struct.int32x2x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <2 x i32>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5126,8 +5126,8 @@ int32x2x2_t test_vld2_lane_s32(int32_t const * a, int32x2x2_t b) {
// CHECK-LABEL: @test_vld2_lane_f16(
// 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: [[__S1:%.*]] = alloca %struct.float16x4x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x2_t, %struct.float16x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x half>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5153,8 +5153,8 @@ float16x4x2_t test_vld2_lane_f16(float16_t const * a, float16x4x2_t b) {
// CHECK-LABEL: @test_vld2_lane_f32(
// 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: [[__S1:%.*]] = alloca %struct.float32x2x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x2_t, %struct.float32x2x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <2 x float>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5180,8 +5180,8 @@ float32x2x2_t test_vld2_lane_f32(float32_t const * a, float32x2x2_t b) {
// CHECK-LABEL: @test_vld2_lane_p8(
// 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: [[__S1:%.*]] = alloca %struct.poly8x8x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x2_t, %struct.poly8x8x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <8 x i8>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5202,8 +5202,8 @@ poly8x8x2_t test_vld2_lane_p8(poly8_t const * a, poly8x8x2_t b) {
// CHECK-LABEL: @test_vld2_lane_p16(
// 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: [[__S1:%.*]] = alloca %struct.poly16x4x2_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x2_t, %struct.poly16x4x2_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [2 x <4 x i16>]* [[COERCE_DIVE]] to [2 x i64]*
// CHECK: store [2 x i64] [[B]].coerce, [2 x i64]* [[TMP0]], align 8
@@ -5421,8 +5421,8 @@ poly16x4x3_t test_vld3_p16(poly16_t const * a) {
// CHECK-LABEL: @test_vld3q_lane_u16(
// 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: [[__S1:%.*]] = alloca %struct.uint16x8x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x3_t, %struct.uint16x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x i16>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5453,8 +5453,8 @@ uint16x8x3_t test_vld3q_lane_u16(uint16_t const * a, uint16x8x3_t b) {
// CHECK-LABEL: @test_vld3q_lane_u32(
// 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: [[__S1:%.*]] = alloca %struct.uint32x4x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x3_t, %struct.uint32x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x i32>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5485,8 +5485,8 @@ uint32x4x3_t test_vld3q_lane_u32(uint32_t const * a, uint32x4x3_t b) {
// CHECK-LABEL: @test_vld3q_lane_s16(
// 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: [[__S1:%.*]] = alloca %struct.int16x8x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x3_t, %struct.int16x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x i16>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5517,8 +5517,8 @@ int16x8x3_t test_vld3q_lane_s16(int16_t const * a, int16x8x3_t b) {
// CHECK-LABEL: @test_vld3q_lane_s32(
// 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: [[__S1:%.*]] = alloca %struct.int32x4x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x3_t, %struct.int32x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x i32>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5549,8 +5549,8 @@ int32x4x3_t test_vld3q_lane_s32(int32_t const * a, int32x4x3_t b) {
// CHECK-LABEL: @test_vld3q_lane_f16(
// 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: [[__S1:%.*]] = alloca %struct.float16x8x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x3_t, %struct.float16x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x half>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5581,8 +5581,8 @@ float16x8x3_t test_vld3q_lane_f16(float16_t const * a, float16x8x3_t b) {
// CHECK-LABEL: @test_vld3q_lane_f32(
// 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: [[__S1:%.*]] = alloca %struct.float32x4x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x3_t, %struct.float32x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x float>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5613,8 +5613,8 @@ float32x4x3_t test_vld3q_lane_f32(float32_t const * a, float32x4x3_t b) {
// CHECK-LABEL: @test_vld3q_lane_p16(
// 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: [[__S1:%.*]] = alloca %struct.poly16x8x3_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x3_t, %struct.poly16x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x i16>]* [[COERCE_DIVE]] to [6 x i64]*
// CHECK: store [6 x i64] [[B]].coerce, [6 x i64]* [[TMP0]], align 16
@@ -5645,8 +5645,8 @@ poly16x8x3_t test_vld3q_lane_p16(poly16_t const * a, poly16x8x3_t b) {
// CHECK-LABEL: @test_vld3_lane_u8(
// 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: [[__S1:%.*]] = alloca %struct.uint8x8x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x3_t, %struct.uint8x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x i8>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5670,8 +5670,8 @@ uint8x8x3_t test_vld3_lane_u8(uint8_t const * a, uint8x8x3_t b) {
// CHECK-LABEL: @test_vld3_lane_u16(
// 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: [[__S1:%.*]] = alloca %struct.uint16x4x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x3_t, %struct.uint16x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x i16>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5702,8 +5702,8 @@ uint16x4x3_t test_vld3_lane_u16(uint16_t const * a, uint16x4x3_t b) {
// CHECK-LABEL: @test_vld3_lane_u32(
// 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: [[__S1:%.*]] = alloca %struct.uint32x2x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x3_t, %struct.uint32x2x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <2 x i32>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5734,8 +5734,8 @@ uint32x2x3_t test_vld3_lane_u32(uint32_t const * a, uint32x2x3_t b) {
// CHECK-LABEL: @test_vld3_lane_s8(
// 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: [[__S1:%.*]] = alloca %struct.int8x8x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x3_t, %struct.int8x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x i8>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5759,8 +5759,8 @@ int8x8x3_t test_vld3_lane_s8(int8_t const * a, int8x8x3_t b) {
// CHECK-LABEL: @test_vld3_lane_s16(
// 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: [[__S1:%.*]] = alloca %struct.int16x4x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x3_t, %struct.int16x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x i16>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5791,8 +5791,8 @@ int16x4x3_t test_vld3_lane_s16(int16_t const * a, int16x4x3_t b) {
// CHECK-LABEL: @test_vld3_lane_s32(
// 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: [[__S1:%.*]] = alloca %struct.int32x2x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x3_t, %struct.int32x2x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <2 x i32>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5823,8 +5823,8 @@ int32x2x3_t test_vld3_lane_s32(int32_t const * a, int32x2x3_t b) {
// CHECK-LABEL: @test_vld3_lane_f16(
// 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: [[__S1:%.*]] = alloca %struct.float16x4x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x3_t, %struct.float16x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x half>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5855,8 +5855,8 @@ float16x4x3_t test_vld3_lane_f16(float16_t const * a, float16x4x3_t b) {
// CHECK-LABEL: @test_vld3_lane_f32(
// 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: [[__S1:%.*]] = alloca %struct.float32x2x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x3_t, %struct.float32x2x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <2 x float>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5887,8 +5887,8 @@ float32x2x3_t test_vld3_lane_f32(float32_t const * a, float32x2x3_t b) {
// CHECK-LABEL: @test_vld3_lane_p8(
// 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: [[__S1:%.*]] = alloca %struct.poly8x8x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x3_t, %struct.poly8x8x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <8 x i8>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -5912,8 +5912,8 @@ poly8x8x3_t test_vld3_lane_p8(poly8_t const * a, poly8x8x3_t b) {
// CHECK-LABEL: @test_vld3_lane_p16(
// 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: [[__S1:%.*]] = alloca %struct.poly16x4x3_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x3_t, %struct.poly16x4x3_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [3 x <4 x i16>]* [[COERCE_DIVE]] to [3 x i64]*
// CHECK: store [3 x i64] [[B]].coerce, [3 x i64]* [[TMP0]], align 8
@@ -6136,8 +6136,8 @@ poly16x4x4_t test_vld4_p16(poly16_t const * a) {
// CHECK-LABEL: @test_vld4q_lane_u16(
// 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: [[__S1:%.*]] = alloca %struct.uint16x8x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x8x4_t, %struct.uint16x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x i16>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6173,8 +6173,8 @@ uint16x8x4_t test_vld4q_lane_u16(uint16_t const * a, uint16x8x4_t b) {
// CHECK-LABEL: @test_vld4q_lane_u32(
// 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: [[__S1:%.*]] = alloca %struct.uint32x4x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x4x4_t, %struct.uint32x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x i32>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6210,8 +6210,8 @@ uint32x4x4_t test_vld4q_lane_u32(uint32_t const * a, uint32x4x4_t b) {
// CHECK-LABEL: @test_vld4q_lane_s16(
// 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: [[__S1:%.*]] = alloca %struct.int16x8x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x8x4_t, %struct.int16x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x i16>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6247,8 +6247,8 @@ int16x8x4_t test_vld4q_lane_s16(int16_t const * a, int16x8x4_t b) {
// CHECK-LABEL: @test_vld4q_lane_s32(
// 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: [[__S1:%.*]] = alloca %struct.int32x4x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x4x4_t, %struct.int32x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x i32>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6284,8 +6284,8 @@ int32x4x4_t test_vld4q_lane_s32(int32_t const * a, int32x4x4_t b) {
// CHECK-LABEL: @test_vld4q_lane_f16(
// 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: [[__S1:%.*]] = alloca %struct.float16x8x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x8x4_t, %struct.float16x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x half>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6321,8 +6321,8 @@ float16x8x4_t test_vld4q_lane_f16(float16_t const * a, float16x8x4_t b) {
// CHECK-LABEL: @test_vld4q_lane_f32(
// 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: [[__S1:%.*]] = alloca %struct.float32x4x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x4x4_t, %struct.float32x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x float>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6358,8 +6358,8 @@ float32x4x4_t test_vld4q_lane_f32(float32_t const * a, float32x4x4_t b) {
// CHECK-LABEL: @test_vld4q_lane_p16(
// 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: [[__S1:%.*]] = alloca %struct.poly16x8x4_t, align 16
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x8x4_t, %struct.poly16x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x i16>]* [[COERCE_DIVE]] to [8 x i64]*
// CHECK: store [8 x i64] [[B]].coerce, [8 x i64]* [[TMP0]], align 16
@@ -6395,8 +6395,8 @@ poly16x8x4_t test_vld4q_lane_p16(poly16_t const * a, poly16x8x4_t b) {
// CHECK-LABEL: @test_vld4_lane_u8(
// 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: [[__S1:%.*]] = alloca %struct.uint8x8x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint8x8x4_t, %struct.uint8x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x i8>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6423,8 +6423,8 @@ uint8x8x4_t test_vld4_lane_u8(uint8_t const * a, uint8x8x4_t b) {
// CHECK-LABEL: @test_vld4_lane_u16(
// 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: [[__S1:%.*]] = alloca %struct.uint16x4x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint16x4x4_t, %struct.uint16x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x i16>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6460,8 +6460,8 @@ uint16x4x4_t test_vld4_lane_u16(uint16_t const * a, uint16x4x4_t b) {
// CHECK-LABEL: @test_vld4_lane_u32(
// 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: [[__S1:%.*]] = alloca %struct.uint32x2x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.uint32x2x4_t, %struct.uint32x2x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <2 x i32>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6497,8 +6497,8 @@ uint32x2x4_t test_vld4_lane_u32(uint32_t const * a, uint32x2x4_t b) {
// CHECK-LABEL: @test_vld4_lane_s8(
// 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: [[__S1:%.*]] = alloca %struct.int8x8x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int8x8x4_t, %struct.int8x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x i8>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6525,8 +6525,8 @@ int8x8x4_t test_vld4_lane_s8(int8_t const * a, int8x8x4_t b) {
// CHECK-LABEL: @test_vld4_lane_s16(
// 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: [[__S1:%.*]] = alloca %struct.int16x4x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int16x4x4_t, %struct.int16x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x i16>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6562,8 +6562,8 @@ int16x4x4_t test_vld4_lane_s16(int16_t const * a, int16x4x4_t b) {
// CHECK-LABEL: @test_vld4_lane_s32(
// 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: [[__S1:%.*]] = alloca %struct.int32x2x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.int32x2x4_t, %struct.int32x2x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <2 x i32>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6599,8 +6599,8 @@ int32x2x4_t test_vld4_lane_s32(int32_t const * a, int32x2x4_t b) {
// CHECK-LABEL: @test_vld4_lane_f16(
// 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: [[__S1:%.*]] = alloca %struct.float16x4x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float16x4x4_t, %struct.float16x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x half>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6636,8 +6636,8 @@ float16x4x4_t test_vld4_lane_f16(float16_t const * a, float16x4x4_t b) {
// CHECK-LABEL: @test_vld4_lane_f32(
// 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: [[__S1:%.*]] = alloca %struct.float32x2x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.float32x2x4_t, %struct.float32x2x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <2 x float>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6673,8 +6673,8 @@ float32x2x4_t test_vld4_lane_f32(float32_t const * a, float32x2x4_t b) {
// CHECK-LABEL: @test_vld4_lane_p8(
// 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: [[__S1:%.*]] = alloca %struct.poly8x8x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly8x8x4_t, %struct.poly8x8x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <8 x i8>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
@@ -6701,8 +6701,8 @@ poly8x8x4_t test_vld4_lane_p8(poly8_t const * a, poly8x8x4_t b) {
// CHECK-LABEL: @test_vld4_lane_p16(
// 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: [[__S1:%.*]] = alloca %struct.poly16x4x4_t, align 8
// CHECK: [[COERCE_DIVE:%.*]] = getelementptr inbounds %struct.poly16x4x4_t, %struct.poly16x4x4_t* [[B]], i32 0, i32 0
// CHECK: [[TMP0:%.*]] = bitcast [4 x <4 x i16>]* [[COERCE_DIVE]] to [4 x i64]*
// CHECK: store [4 x i64] [[B]].coerce, [4 x i64]* [[TMP0]], align 8
diff --git a/clang/test/Sema/arm-neon-decl-after-stmt.c b/clang/test/Sema/arm-neon-decl-after-stmt.c
new file mode 100644
index 0000000000000..c6529445c6f5b
--- /dev/null
+++ b/clang/test/Sema/arm-neon-decl-after-stmt.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -triple armebv7-linux-gnueabihf -target-feature +neon \
+// RUN: -Wdeclaration-after-statement -fsyntax-only -verify %s
+// REQUIRES: arm-registered-target
+// https://github.com/llvm/llvm-project/issues/54062
+#include <arm_neon.h>
+
+uint8x16_t a;
+
+uint8x16_t x(void) {
+ return vshrq_n_u8(a, 8);
+}
+// expected-no-diagnostics
diff --git a/clang/utils/TableGen/NeonEmitter.cpp b/clang/utils/TableGen/NeonEmitter.cpp
index a69fbe0e42dcd..1550a8a1e273e 100644
--- a/clang/utils/TableGen/NeonEmitter.cpp
+++ b/clang/utils/TableGen/NeonEmitter.cpp
@@ -502,6 +502,7 @@ class Intrinsic {
void emitBody(StringRef CallPrefix);
void emitShadowedArgs();
void emitArgumentReversal();
+ void emitReturnVarDecl();
void emitReturnReversal();
void emitReverseVariable(Variable &Dest, Variable &Src);
void emitNewLine();
@@ -1228,6 +1229,15 @@ void Intrinsic::emitArgumentReversal() {
}
}
+void Intrinsic::emitReturnVarDecl() {
+ assert(RetVar.getType() == Types[0]);
+ // Create a return variable, if we're not void.
+ if (!RetVar.getType().isVoid()) {
+ OS << " " << RetVar.getType().str() << " " << RetVar.getName() << ";";
+ emitNewLine();
+ }
+}
+
void Intrinsic::emitReturnReversal() {
if (isBigEndianSafe())
return;
@@ -1353,13 +1363,6 @@ void Intrinsic::emitBodyAsBuiltinCall() {
void Intrinsic::emitBody(StringRef CallPrefix) {
std::vector<std::string> Lines;
- assert(RetVar.getType() == Types[0]);
- // Create a return variable, if we're not void.
- if (!RetVar.getType().isVoid()) {
- OS << " " << RetVar.getType().str() << " " << RetVar.getName() << ";";
- emitNewLine();
- }
-
if (!Body || Body->getValues().empty()) {
// Nothing specific to output - must output a builtin.
emitBodyAsBuiltinCall();
@@ -1849,6 +1852,9 @@ void Intrinsic::generateImpl(bool ReverseArguments,
OS << " __attribute__((unavailable));";
} else {
emitOpeningBrace();
+ // Emit return variable declaration first as to not trigger
+ // -Wdeclaration-after-statement.
+ emitReturnVarDecl();
emitShadowedArgs();
if (ReverseArguments)
emitArgumentReversal();
@@ -1867,6 +1873,9 @@ void Intrinsic::indexBody() {
CurrentRecord = R;
initVariables();
+ // Emit return variable declaration first as to not trigger
+ // -Wdeclaration-after-statement.
+ emitReturnVarDecl();
emitBody("");
OS.str("");
More information about the cfe-commits
mailing list