// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-attributes --include-generated-funcs --version 4
// RUN: %clang_cc1 -O0 -triple i686-pc-darwin -emit-llvm -o - %s | FileCheck -check-prefix=X86 %s
// RUN: %clang_cc1 -O0 -triple amdgcn -emit-llvm -o - %s | FileCheck -check-prefix=AMDGCN %s

#pragma OPENCL EXTENSION __cl_clang_function_pointers : enable

typedef int int2 __attribute__((ext_vector_type(2)));

typedef struct {
  int cells[9];
} Mat3X3;

typedef struct {
  int cells[16];
} Mat4X4;

typedef struct {
  int cells[1024];
} Mat32X32;

typedef struct {
  int cells[4096];
} Mat64X64;

struct StructOneMember {
  int2 x;
};

struct StructTwoMember {
  int2 x;
  int2 y;
};

struct LargeStructOneMember {
  int2 x[100];
};

struct LargeStructTwoMember {
  int2 x[40];
  int2 y[20];
};

Mat4X4 __attribute__((noinline)) foo(Mat3X3 in) {
  Mat4X4 out;
  return out;
}

Mat64X64 __attribute__((noinline)) foo_large(Mat32X32 in) {
  Mat64X64 out;
  return out;
}

void FuncOneMember(struct StructOneMember u) {
  u.x = (int2)(0, 0);
}

void FuncOneLargeMember(struct LargeStructOneMember u) {
  u.x[0] = (int2)(0, 0);
}

void FuncTwoMember(struct StructTwoMember u) {
  u.y = (int2)(0, 0);
}

void FuncLargeTwoMember(struct LargeStructTwoMember u) {
  u.y[0] = (int2)(0, 0);
}

__attribute__((noinline)) kernel void callee_kern(global int *A){
  *A = 1;
}

__attribute__((optnone)) kernel void callee_kern_with_optnone_attribute(global int *A){
  *A = 1;
}

__attribute__((always_inline)) kernel void callee_kern_with_alwaysinline_attribute(global int *A){
  *A = 1;
}

kernel void callee_kern_Mat3X3(global Mat3X3 *in, global Mat4X4 *out) {
  out[0] = foo(in[1]);
}

kernel void callee_kern_Mat32X32(global Mat32X32 *in, global Mat64X64 *out) {
  out[0] = foo_large(in[1]);
}

kernel void KernelOneMember(struct StructOneMember u) {
  FuncOneMember(u);
}

kernel void KernelLargeOneMember(struct LargeStructOneMember u) {
  FuncOneLargeMember(u);
}

kernel void KernelTwoMember(struct StructTwoMember u) {
  FuncTwoMember(u);
}

kernel void KernelLargeTwoMember(struct LargeStructTwoMember u) {
  FuncLargeTwoMember(u);
}

__attribute__((noinline)) kernel void ext_callee_kern(global int *A);

kernel void ext_callee_kern_Mat3X3(global Mat3X3 *in, global Mat4X4 *out);

kernel void ext_callee_kern_Mat32X32(global Mat32X32 *in, global Mat64X64 *out);

kernel void ext_KernelOneMember(struct StructOneMember u);

kernel void ext_KernelLargeOneMember(struct LargeStructOneMember u);

kernel void ext_KernelTwoMember(struct StructTwoMember u);

kernel void ext_KernelLargeTwoMember(struct LargeStructTwoMember u);

kernel void caller_kern(global int* A, global Mat3X3 *mat3X3, global Mat4X4 *mat4X4, global Mat32X32 *mat32X32, global Mat64X64 *mat64X64){
  callee_kern(A);
  ext_callee_kern(A);
  callee_kern_with_optnone_attribute(A);
  callee_kern_with_alwaysinline_attribute(A);
  callee_kern_Mat3X3(mat3X3, mat4X4);
  callee_kern_Mat32X32(mat32X32, mat64X64);
  ext_callee_kern_Mat3X3(mat3X3, mat4X4);
  ext_callee_kern_Mat32X32(mat32X32, mat64X64);
}

kernel void caller_kern2(struct StructOneMember structOneMem, global struct StructOneMember* global_structOneMem, struct StructTwoMember structTwoMem){
  KernelOneMember(structOneMem);
  ext_KernelOneMember(structOneMem);
  KernelTwoMember(structTwoMem);
  ext_KernelTwoMember(structTwoMem);
}

kernel void caller_kern3( struct LargeStructOneMember largeStructOneMem, struct LargeStructTwoMember largeStructTwoMem){
  KernelLargeOneMember(largeStructOneMem);
  KernelLargeTwoMember(largeStructTwoMem);
  ext_KernelLargeOneMember(largeStructOneMem);
  ext_KernelLargeTwoMember(largeStructTwoMem);
}
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @foo(
// X86-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_MAT4X4:%.*]]) align 4 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_MAT3X3:%.*]]) align 4 [[IN:%.*]]) #[[ATTR0:[0-9]+]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[RESULT_PTR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @foo_large(
// X86-SAME: ptr dead_on_unwind noalias writable sret([[STRUCT_MAT64X64:%.*]]) align 4 [[AGG_RESULT:%.*]], ptr noundef byval([[STRUCT_MAT32X32:%.*]]) align 4 [[IN:%.*]]) #[[ATTR0]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[RESULT_PTR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[AGG_RESULT]], ptr [[RESULT_PTR]], align 4
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @FuncOneMember(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTONEMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER]], align 8
// X86-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 8, i1 false)
// X86-NEXT:    store <2 x i32> zeroinitializer, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr [[U]], i32 0, i32 0
// X86-NEXT:    store <2 x i32> [[TMP1]], ptr [[X]], align 8
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @FuncOneLargeMember(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8
// X86-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 800, i1 false)
// X86-NEXT:    store <2 x i32> zeroinitializer, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_LARGESTRUCTONEMEMBER]], ptr [[U]], i32 0, i32 0
// X86-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x <2 x i32>], ptr [[X]], i32 0, i32 0
// X86-NEXT:    store <2 x i32> [[TMP1]], ptr [[ARRAYIDX]], align 8
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @FuncTwoMember(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER]], align 8
// X86-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 16, i1 false)
// X86-NEXT:    store <2 x i32> zeroinitializer, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr [[U]], i32 0, i32 1
// X86-NEXT:    store <2 x i32> [[TMP1]], ptr [[Y]], align 8
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @FuncLargeTwoMember(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8
// X86-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 480, i1 false)
// X86-NEXT:    store <2 x i32> zeroinitializer, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr [[DOTCOMPOUNDLITERAL]], align 8
// X86-NEXT:    [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_LARGESTRUCTTWOMEMBER]], ptr [[U]], i32 0, i32 1
// X86-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [20 x <2 x i32>], ptr [[Y]], i32 0, i32 0
// X86-NEXT:    store <2 x i32> [[TMP1]], ptr [[ARRAYIDX]], align 8
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @callee_kern(
// X86-SAME: ptr noundef align 4 [[A:%.*]]) #[[ATTR2:[0-9]+]] !kernel_arg_addr_space [[META4:![0-9]+]] !kernel_arg_access_qual [[META5:![0-9]+]] !kernel_arg_type [[META6:![0-9]+]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7:![0-9]+]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern(ptr noundef align 4 [[TMP0]]) #[[ATTR6:[0-9]+]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_callee_kern(
// X86-SAME: ptr noundef align 4 [[A:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    store i32 1, ptr [[TMP0]], align 4
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @callee_kern_with_optnone_attribute(
// X86-SAME: ptr noundef align 4 [[A:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_with_optnone_attribute(ptr noundef align 4 [[TMP0]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_callee_kern_with_optnone_attribute(
// X86-SAME: ptr noundef align 4 [[A:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    store i32 1, ptr [[TMP0]], align 4
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: alwaysinline convergent norecurse nounwind
// X86-LABEL: define spir_kernel void @callee_kern_with_alwaysinline_attribute(
// X86-SAME: ptr noundef align 4 [[A:%.*]]) #[[ATTR3:[0-9]+]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR_I:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    store ptr [[TMP0]], ptr [[A_ADDR_I]], align 4
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_I]], align 4
// X86-NEXT:    store i32 1, ptr [[TMP1]], align 4
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: alwaysinline convergent norecurse nounwind
// X86-LABEL: define void @__clang_ocl_kern_imp_callee_kern_with_alwaysinline_attribute(
// X86-SAME: ptr noundef align 4 [[A:%.*]]) #[[ATTR4:[0-9]+]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    store i32 1, ptr [[TMP0]], align 4
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @callee_kern_Mat3X3(
// X86-SAME: ptr noundef align 4 [[IN:%.*]], ptr noundef align 4 [[OUT:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META8:![0-9]+]] !kernel_arg_access_qual [[META9:![0-9]+]] !kernel_arg_type [[META10:![0-9]+]] !kernel_arg_base_type [[META10]] !kernel_arg_type_qual [[META11:![0-9]+]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[IN_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[IN]], ptr [[IN_ADDR]], align 4
// X86-NEXT:    store ptr [[OUT]], ptr [[OUT_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[IN_ADDR]], align 4
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[OUT_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat3X3(ptr noundef align 4 [[TMP0]], ptr noundef align 4 [[TMP1]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_callee_kern_Mat3X3(
// X86-SAME: ptr noundef align 4 [[IN:%.*]], ptr noundef align 4 [[OUT:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META8]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META10]] !kernel_arg_base_type [[META10]] !kernel_arg_type_qual [[META11]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[IN_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[TMP:%.*]] = alloca [[STRUCT_MAT4X4:%.*]], align 4
// X86-NEXT:    store ptr [[IN]], ptr [[IN_ADDR]], align 4
// X86-NEXT:    store ptr [[OUT]], ptr [[OUT_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OUT_ADDR]], align 4
// X86-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_MAT4X4]], ptr [[TMP0]], i32 0
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[IN_ADDR]], align 4
// X86-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_MAT3X3:%.*]], ptr [[TMP1]], i32 1
// X86-NEXT:    call void @foo(ptr dead_on_unwind writable sret([[STRUCT_MAT4X4]]) align 4 [[TMP]], ptr noundef byval([[STRUCT_MAT3X3]]) align 4 [[ARRAYIDX1]]) #[[ATTR6]]
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX]], ptr align 4 [[TMP]], i32 64, i1 false)
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @callee_kern_Mat32X32(
// X86-SAME: ptr noundef align 4 [[IN:%.*]], ptr noundef align 4 [[OUT:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META8]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META12:![0-9]+]] !kernel_arg_base_type [[META12]] !kernel_arg_type_qual [[META11]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[IN_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[IN]], ptr [[IN_ADDR]], align 4
// X86-NEXT:    store ptr [[OUT]], ptr [[OUT_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[IN_ADDR]], align 4
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[OUT_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat32X32(ptr noundef align 4 [[TMP0]], ptr noundef align 4 [[TMP1]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_callee_kern_Mat32X32(
// X86-SAME: ptr noundef align 4 [[IN:%.*]], ptr noundef align 4 [[OUT:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META8]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META12]] !kernel_arg_base_type [[META12]] !kernel_arg_type_qual [[META11]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[IN_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[TMP:%.*]] = alloca [[STRUCT_MAT64X64:%.*]], align 4
// X86-NEXT:    store ptr [[IN]], ptr [[IN_ADDR]], align 4
// X86-NEXT:    store ptr [[OUT]], ptr [[OUT_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OUT_ADDR]], align 4
// X86-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_MAT64X64]], ptr [[TMP0]], i32 0
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[IN_ADDR]], align 4
// X86-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_MAT32X32:%.*]], ptr [[TMP1]], i32 1
// X86-NEXT:    call void @foo_large(ptr dead_on_unwind writable sret([[STRUCT_MAT64X64]]) align 4 [[TMP]], ptr noundef byval([[STRUCT_MAT32X32]]) align 4 [[ARRAYIDX1]]) #[[ATTR6]]
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX]], ptr align 4 [[TMP]], i32 16384, i1 false)
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @KernelOneMember(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTONEMEMBER:%.*]]) align 8 [[U:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13:![0-9]+]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META14:![0-9]+]] !kernel_arg_base_type [[META14]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelOneMember(ptr noundef byval([[STRUCT_STRUCTONEMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_KernelOneMember(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTONEMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META14]] !kernel_arg_base_type [[META14]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER]], align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 8, i1 false)
// X86-NEXT:    call void @FuncOneMember(ptr noundef byval([[STRUCT_STRUCTONEMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @KernelLargeOneMember(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[U:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META15:![0-9]+]] !kernel_arg_base_type [[META15]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeOneMember(ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_KernelLargeOneMember(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META15]] !kernel_arg_base_type [[META15]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 800, i1 false)
// X86-NEXT:    call void @FuncOneLargeMember(ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @KernelTwoMember(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 8 [[U:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META16:![0-9]+]] !kernel_arg_base_type [[META16]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelTwoMember(ptr noundef byval([[STRUCT_STRUCTTWOMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_KernelTwoMember(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META16]] !kernel_arg_base_type [[META16]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER]], align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 16, i1 false)
// X86-NEXT:    call void @FuncTwoMember(ptr noundef byval([[STRUCT_STRUCTTWOMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @KernelLargeTwoMember(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[U:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META17:![0-9]+]] !kernel_arg_base_type [[META17]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeTwoMember(ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_KernelLargeTwoMember(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META17]] !kernel_arg_base_type [[META17]] !kernel_arg_type_qual [[META7]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[U]], ptr align 4 [[TMP0]], i32 480, i1 false)
// X86-NEXT:    call void @FuncLargeTwoMember(ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 4 [[U]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @caller_kern(
// X86-SAME: ptr noundef align 4 [[A:%.*]], ptr noundef align 4 [[MAT3X3:%.*]], ptr noundef align 4 [[MAT4X4:%.*]], ptr noundef align 4 [[MAT32X32:%.*]], ptr noundef align 4 [[MAT64X64:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META18:![0-9]+]] !kernel_arg_access_qual [[META19:![0-9]+]] !kernel_arg_type [[META20:![0-9]+]] !kernel_arg_base_type [[META20]] !kernel_arg_type_qual [[META21:![0-9]+]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT3X3_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT4X4_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT32X32_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT64X64_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT3X3]], ptr [[MAT3X3_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT4X4]], ptr [[MAT4X4_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT32X32]], ptr [[MAT32X32_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT64X64]], ptr [[MAT64X64_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[MAT3X3_ADDR]], align 4
// X86-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[MAT4X4_ADDR]], align 4
// X86-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[MAT32X32_ADDR]], align 4
// X86-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[MAT64X64_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_caller_kern(ptr noundef align 4 [[TMP0]], ptr noundef align 4 [[TMP1]], ptr noundef align 4 [[TMP2]], ptr noundef align 4 [[TMP3]], ptr noundef align 4 [[TMP4]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_caller_kern(
// X86-SAME: ptr noundef align 4 [[A:%.*]], ptr noundef align 4 [[MAT3X3:%.*]], ptr noundef align 4 [[MAT4X4:%.*]], ptr noundef align 4 [[MAT32X32:%.*]], ptr noundef align 4 [[MAT64X64:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META18]] !kernel_arg_access_qual [[META19]] !kernel_arg_type [[META20]] !kernel_arg_base_type [[META20]] !kernel_arg_type_qual [[META21]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[A_ADDR_I:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT3X3_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT4X4_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT32X32_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    [[MAT64X64_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT3X3]], ptr [[MAT3X3_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT4X4]], ptr [[MAT4X4_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT32X32]], ptr [[MAT32X32_ADDR]], align 4
// X86-NEXT:    store ptr [[MAT64X64]], ptr [[MAT64X64_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern(ptr noundef align 4 [[TMP0]]) #[[ATTR6]]
// X86-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_callee_kern(ptr noundef align 4 [[TMP1]]) #[[ATTR6]]
// X86-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_with_optnone_attribute(ptr noundef align 4 [[TMP2]]) #[[ATTR6]]
// X86-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[A_ADDR]], align 4
// X86-NEXT:    store ptr [[TMP3]], ptr [[A_ADDR_I]], align 4
// X86-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR_I]], align 4
// X86-NEXT:    store i32 1, ptr [[TMP4]], align 4
// X86-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[MAT3X3_ADDR]], align 4
// X86-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[MAT4X4_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat3X3(ptr noundef align 4 [[TMP5]], ptr noundef align 4 [[TMP6]]) #[[ATTR6]]
// X86-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[MAT32X32_ADDR]], align 4
// X86-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[MAT64X64_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat32X32(ptr noundef align 4 [[TMP7]], ptr noundef align 4 [[TMP8]]) #[[ATTR6]]
// X86-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[MAT3X3_ADDR]], align 4
// X86-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[MAT4X4_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_callee_kern_Mat3X3(ptr noundef align 4 [[TMP9]], ptr noundef align 4 [[TMP10]]) #[[ATTR6]]
// X86-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[MAT32X32_ADDR]], align 4
// X86-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[MAT64X64_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_callee_kern_Mat32X32(ptr noundef align 4 [[TMP11]], ptr noundef align 4 [[TMP12]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @caller_kern2(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTONEMEMBER:%.*]]) align 8 [[STRUCTONEMEM:%.*]], ptr noundef align 8 [[GLOBAL_STRUCTONEMEM:%.*]], ptr noundef byval([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 8 [[STRUCTTWOMEM:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META22:![0-9]+]] !kernel_arg_access_qual [[META23:![0-9]+]] !kernel_arg_type [[META24:![0-9]+]] !kernel_arg_base_type [[META24]] !kernel_arg_type_qual [[META25:![0-9]+]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[GLOBAL_STRUCTONEMEM_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    store ptr [[GLOBAL_STRUCTONEMEM]], ptr [[GLOBAL_STRUCTONEMEM_ADDR]], align 4
// X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[GLOBAL_STRUCTONEMEM_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_caller_kern2(ptr noundef byval([[STRUCT_STRUCTONEMEMBER]]) align 4 [[STRUCTONEMEM]], ptr noundef align 8 [[TMP0]], ptr noundef byval([[STRUCT_STRUCTTWOMEMBER]]) align 4 [[STRUCTTWOMEM]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_caller_kern2(
// X86-SAME: ptr noundef byval([[STRUCT_STRUCTONEMEMBER:%.*]]) align 4 [[TMP0:%.*]], ptr noundef align 8 [[GLOBAL_STRUCTONEMEM:%.*]], ptr noundef byval([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 4 [[TMP1:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META22]] !kernel_arg_access_qual [[META23]] !kernel_arg_type [[META24]] !kernel_arg_base_type [[META24]] !kernel_arg_type_qual [[META25]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[STRUCTONEMEM:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER]], align 8
// X86-NEXT:    [[STRUCTTWOMEM:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER]], align 8
// X86-NEXT:    [[GLOBAL_STRUCTONEMEM_ADDR:%.*]] = alloca ptr, align 4
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[STRUCTONEMEM]], ptr align 4 [[TMP0]], i32 8, i1 false)
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[STRUCTTWOMEM]], ptr align 4 [[TMP1]], i32 16, i1 false)
// X86-NEXT:    store ptr [[GLOBAL_STRUCTONEMEM]], ptr [[GLOBAL_STRUCTONEMEM_ADDR]], align 4
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelOneMember(ptr noundef byval([[STRUCT_STRUCTONEMEMBER]]) align 4 [[STRUCTONEMEM]]) #[[ATTR6]]
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelOneMember(ptr noundef byval([[STRUCT_STRUCTONEMEMBER]]) align 4 [[STRUCTONEMEM]]) #[[ATTR6]]
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelTwoMember(ptr noundef byval([[STRUCT_STRUCTTWOMEMBER]]) align 4 [[STRUCTTWOMEM]]) #[[ATTR6]]
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelTwoMember(ptr noundef byval([[STRUCT_STRUCTTWOMEMBER]]) align 4 [[STRUCTTWOMEM]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define spir_kernel void @caller_kern3(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[LARGESTRUCTONEMEM:%.*]], ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[LARGESTRUCTTWOMEM:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META26:![0-9]+]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META27:![0-9]+]] !kernel_arg_base_type [[META27]] !kernel_arg_type_qual [[META11]] {
// X86-NEXT:  entry:
// X86-NEXT:    call void @__clang_ocl_kern_imp_caller_kern3(ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER]]) align 4 [[LARGESTRUCTONEMEM]], ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 4 [[LARGESTRUCTTWOMEM]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// X86: Function Attrs: convergent noinline norecurse nounwind optnone
// X86-LABEL: define void @__clang_ocl_kern_imp_caller_kern3(
// X86-SAME: ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 4 [[TMP0:%.*]], ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 4 [[TMP1:%.*]]) #[[ATTR0]] !kernel_arg_addr_space [[META26]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META27]] !kernel_arg_base_type [[META27]] !kernel_arg_type_qual [[META11]] {
// X86-NEXT:  entry:
// X86-NEXT:    [[LARGESTRUCTONEMEM:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8
// X86-NEXT:    [[LARGESTRUCTTWOMEM:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[LARGESTRUCTONEMEM]], ptr align 4 [[TMP0]], i32 800, i1 false)
// X86-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 8 [[LARGESTRUCTTWOMEM]], ptr align 4 [[TMP1]], i32 480, i1 false)
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeOneMember(ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER]]) align 4 [[LARGESTRUCTONEMEM]]) #[[ATTR6]]
// X86-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeTwoMember(ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 4 [[LARGESTRUCTTWOMEM]]) #[[ATTR6]]
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelLargeOneMember(ptr noundef byval([[STRUCT_LARGESTRUCTONEMEMBER]]) align 4 [[LARGESTRUCTONEMEM]]) #[[ATTR6]]
// X86-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelLargeTwoMember(ptr noundef byval([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 4 [[LARGESTRUCTTWOMEM]]) #[[ATTR6]]
// X86-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local %struct.Mat4X4 @foo(
// AMDGCN-SAME: [9 x i32] [[IN_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[RETVAL:%.*]] = alloca [[STRUCT_MAT4X4:%.*]], align 4, addrspace(5)
// AMDGCN-NEXT:    [[IN:%.*]] = alloca [[STRUCT_MAT3X3:%.*]], align 4, addrspace(5)
// AMDGCN-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_MAT3X3]], ptr addrspace(5) [[IN]], i32 0, i32 0
// AMDGCN-NEXT:    store [9 x i32] [[IN_COERCE]], ptr addrspace(5) [[COERCE_DIVE]], align 4
// AMDGCN-NEXT:    [[TMP0:%.*]] = load [[STRUCT_MAT4X4]], ptr addrspace(5) [[RETVAL]], align 4
// AMDGCN-NEXT:    ret [[STRUCT_MAT4X4]] [[TMP0]]
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @foo_large(
// AMDGCN-SAME: ptr addrspace(5) dead_on_unwind noalias writable sret([[STRUCT_MAT64X64:%.*]]) align 4 [[AGG_RESULT:%.*]], ptr addrspace(5) noundef byref([[STRUCT_MAT32X32:%.*]]) align 4 [[TMP0:%.*]]) #[[ATTR0]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[IN:%.*]] = alloca [[STRUCT_MAT32X32]], align 4, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 4 [[IN]], ptr addrspace(5) align 4 [[TMP0]], i64 4096, i1 false)
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @FuncOneMember(
// AMDGCN-SAME: <2 x i32> [[U_COERCE:%.*]]) #[[ATTR0]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8, addrspace(5)
// AMDGCN-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE]], ptr addrspace(5) [[COERCE_DIVE]], align 8
// AMDGCN-NEXT:    store <2 x i32> zeroinitializer, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[TMP0]], ptr addrspace(5) [[X]], align 8
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @FuncOneLargeMember(
// AMDGCN-SAME: ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR0]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(5) align 8 [[TMP0]], i64 800, i1 false)
// AMDGCN-NEXT:    store <2 x i32> zeroinitializer, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[X:%.*]] = getelementptr inbounds nuw [[STRUCT_LARGESTRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x <2 x i32>], ptr addrspace(5) [[X]], i64 0, i64 0
// AMDGCN-NEXT:    store <2 x i32> [[TMP1]], ptr addrspace(5) [[ARRAYIDX]], align 8
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @FuncTwoMember(
// AMDGCN-SAME: <2 x i32> [[U_COERCE0:%.*]], <2 x i32> [[U_COERCE1:%.*]]) #[[ATTR0]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8, addrspace(5)
// AMDGCN-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE0]], ptr addrspace(5) [[TMP0]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 1
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE1]], ptr addrspace(5) [[TMP1]], align 8
// AMDGCN-NEXT:    store <2 x i32> zeroinitializer, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 1
// AMDGCN-NEXT:    store <2 x i32> [[TMP2]], ptr addrspace(5) [[Y]], align 8
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @FuncLargeTwoMember(
// AMDGCN-SAME: ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR0]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[DOTCOMPOUNDLITERAL:%.*]] = alloca <2 x i32>, align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(5) align 8 [[TMP0]], i64 480, i1 false)
// AMDGCN-NEXT:    store <2 x i32> zeroinitializer, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load <2 x i32>, ptr addrspace(5) [[DOTCOMPOUNDLITERAL]], align 8
// AMDGCN-NEXT:    [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_LARGESTRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 1
// AMDGCN-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [20 x <2 x i32>], ptr addrspace(5) [[Y]], i64 0, i64 0
// AMDGCN-NEXT:    store <2 x i32> [[TMP1]], ptr addrspace(5) [[ARRAYIDX]], align 8
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @callee_kern(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]]) #[[ATTR2:[0-9]+]] !kernel_arg_addr_space [[META4:![0-9]+]] !kernel_arg_access_qual [[META5:![0-9]+]] !kernel_arg_type [[META6:![0-9]+]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7:![0-9]+]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern(ptr addrspace(1) noundef align 4 [[TMP0]]) #[[ATTR7:[0-9]+]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_callee_kern(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]]) #[[ATTR3:[0-9]+]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store i32 1, ptr addrspace(1) [[TMP0]], align 4
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @callee_kern_with_optnone_attribute(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_with_optnone_attribute(ptr addrspace(1) noundef align 4 [[TMP0]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_callee_kern_with_optnone_attribute(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store i32 1, ptr addrspace(1) [[TMP0]], align 4
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: alwaysinline convergent norecurse nounwind
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @callee_kern_with_alwaysinline_attribute(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]]) #[[ATTR4:[0-9]+]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR_I:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[TMP0]], ptr addrspace(5) [[A_ADDR_I]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR_I]], align 8
// AMDGCN-NEXT:    store i32 1, ptr addrspace(1) [[TMP1]], align 4
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: alwaysinline convergent norecurse nounwind
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_callee_kern_with_alwaysinline_attribute(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]]) #[[ATTR5:[0-9]+]] !kernel_arg_addr_space [[META4]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META6]] !kernel_arg_base_type [[META6]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store i32 1, ptr addrspace(1) [[TMP0]], align 4
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @callee_kern_Mat3X3(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[IN:%.*]], ptr addrspace(1) noundef align 4 [[OUT:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META8:![0-9]+]] !kernel_arg_access_qual [[META9:![0-9]+]] !kernel_arg_type [[META10:![0-9]+]] !kernel_arg_base_type [[META10]] !kernel_arg_type_qual [[META11:![0-9]+]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[IN_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[IN]], ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[OUT]], ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat3X3(ptr addrspace(1) noundef align 4 [[TMP0]], ptr addrspace(1) noundef align 4 [[TMP1]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_callee_kern_Mat3X3(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[IN:%.*]], ptr addrspace(1) noundef align 4 [[OUT:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META8]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META10]] !kernel_arg_base_type [[META10]] !kernel_arg_type_qual [[META11]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[IN_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[TMP:%.*]] = alloca [[STRUCT_MAT4X4:%.*]], align 4, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[IN]], ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[OUT]], ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_MAT4X4]], ptr addrspace(1) [[TMP0]], i64 0
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_MAT3X3:%.*]], ptr addrspace(1) [[TMP1]], i64 1
// AMDGCN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT_MAT3X3]], ptr addrspace(1) [[ARRAYIDX1]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP3:%.*]] = load [9 x i32], ptr addrspace(1) [[TMP2]], align 4
// AMDGCN-NEXT:    [[CALL:%.*]] = call [[STRUCT_MAT4X4]] @[[FOO:[a-zA-Z0-9_$\"\\.-]*[a-zA-Z_$\"\\.-][a-zA-Z0-9_$\"\\.-]*]]([9 x i32] [[TMP3]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT_MAT4X4]], ptr addrspace(5) [[TMP]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP5:%.*]] = extractvalue [[STRUCT_MAT4X4]] [[CALL]], 0
// AMDGCN-NEXT:    store [16 x i32] [[TMP5]], ptr addrspace(5) [[TMP4]], align 4
// AMDGCN-NEXT:    call void @llvm.memcpy.p1.p5.i64(ptr addrspace(1) align 4 [[ARRAYIDX]], ptr addrspace(5) align 4 [[TMP]], i64 64, i1 false)
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @callee_kern_Mat32X32(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[IN:%.*]], ptr addrspace(1) noundef align 4 [[OUT:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META8]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META12:![0-9]+]] !kernel_arg_base_type [[META12]] !kernel_arg_type_qual [[META11]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[IN_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[IN]], ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[OUT]], ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat32X32(ptr addrspace(1) noundef align 4 [[TMP0]], ptr addrspace(1) noundef align 4 [[TMP1]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_callee_kern_Mat32X32(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[IN:%.*]], ptr addrspace(1) noundef align 4 [[OUT:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META8]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META12]] !kernel_arg_base_type [[META12]] !kernel_arg_type_qual [[META11]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[IN_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[TMP:%.*]] = alloca [[STRUCT_MAT64X64:%.*]], align 4, addrspace(5)
// AMDGCN-NEXT:    [[BYVAL_TEMP:%.*]] = alloca [[STRUCT_MAT32X32:%.*]], align 4, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[IN]], ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[OUT]], ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[OUT_ADDR]], align 8
// AMDGCN-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [[STRUCT_MAT64X64]], ptr addrspace(1) [[TMP0]], i64 0
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[IN_ADDR]], align 8
// AMDGCN-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_MAT32X32]], ptr addrspace(1) [[TMP1]], i64 1
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p1.i64(ptr addrspace(5) align 4 [[BYVAL_TEMP]], ptr addrspace(1) align 4 [[ARRAYIDX1]], i64 4096, i1 false)
// AMDGCN-NEXT:    call void @foo_large(ptr addrspace(5) dead_on_unwind writable sret([[STRUCT_MAT64X64]]) align 4 [[TMP]], ptr addrspace(5) noundef byref([[STRUCT_MAT32X32]]) align 4 [[BYVAL_TEMP]]) #[[ATTR7]]
// AMDGCN-NEXT:    call void @llvm.memcpy.p1.p5.i64(ptr addrspace(1) align 4 [[ARRAYIDX]], ptr addrspace(5) align 4 [[TMP]], i64 16384, i1 false)
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @KernelOneMember(
// AMDGCN-SAME: <2 x i32> [[U_COERCE:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13:![0-9]+]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META14:![0-9]+]] !kernel_arg_base_type [[META14]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE]], ptr addrspace(5) [[COERCE_DIVE]], align 8
// AMDGCN-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr addrspace(5) [[COERCE_DIVE1]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelOneMember(<2 x i32> [[TMP0]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_KernelOneMember(
// AMDGCN-SAME: <2 x i32> [[U_COERCE:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META14]] !kernel_arg_base_type [[META14]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE]], ptr addrspace(5) [[COERCE_DIVE]], align 8
// AMDGCN-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP0:%.*]] = load <2 x i32>, ptr addrspace(5) [[COERCE_DIVE1]], align 8
// AMDGCN-NEXT:    call void @FuncOneMember(<2 x i32> [[TMP0]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @KernelLargeOneMember(
// AMDGCN-SAME: ptr addrspace(4) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META15:![0-9]+]] !kernel_arg_base_type [[META15]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(4) align 8 [[TMP0]], i64 800, i1 false)
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeOneMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER]]) align 8 [[U]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_KernelLargeOneMember(
// AMDGCN-SAME: ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META15]] !kernel_arg_base_type [[META15]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(5) align 8 [[TMP0]], i64 800, i1 false)
// AMDGCN-NEXT:    call void @FuncOneLargeMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER]]) align 8 [[U]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @KernelTwoMember(
// AMDGCN-SAME: ptr addrspace(4) noundef byref([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META16:![0-9]+]] !kernel_arg_base_type [[META16]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(4) align 8 [[TMP0]], i64 16, i1 false)
// AMDGCN-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP1]], align 8
// AMDGCN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 1
// AMDGCN-NEXT:    [[TMP4:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP3]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelTwoMember(<2 x i32> [[TMP2]], <2 x i32> [[TMP4]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_KernelTwoMember(
// AMDGCN-SAME: <2 x i32> [[U_COERCE0:%.*]], <2 x i32> [[U_COERCE1:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META16]] !kernel_arg_base_type [[META16]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE0]], ptr addrspace(5) [[TMP0]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 1
// AMDGCN-NEXT:    store <2 x i32> [[U_COERCE1]], ptr addrspace(5) [[TMP1]], align 8
// AMDGCN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP3:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP2]], align 8
// AMDGCN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[U]], i32 0, i32 1
// AMDGCN-NEXT:    [[TMP5:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP4]], align 8
// AMDGCN-NEXT:    call void @FuncTwoMember(<2 x i32> [[TMP3]], <2 x i32> [[TMP5]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @KernelLargeTwoMember(
// AMDGCN-SAME: ptr addrspace(4) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META17:![0-9]+]] !kernel_arg_base_type [[META17]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(4) align 8 [[TMP0]], i64 480, i1 false)
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeTwoMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 8 [[U]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_KernelLargeTwoMember(
// AMDGCN-SAME: ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META13]] !kernel_arg_access_qual [[META5]] !kernel_arg_type [[META17]] !kernel_arg_base_type [[META17]] !kernel_arg_type_qual [[META7]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[U:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 8 [[U]], ptr addrspace(5) align 8 [[TMP0]], i64 480, i1 false)
// AMDGCN-NEXT:    call void @FuncLargeTwoMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 8 [[U]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @caller_kern(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]], ptr addrspace(1) noundef align 4 [[MAT3X3:%.*]], ptr addrspace(1) noundef align 4 [[MAT4X4:%.*]], ptr addrspace(1) noundef align 4 [[MAT32X32:%.*]], ptr addrspace(1) noundef align 4 [[MAT64X64:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META18:![0-9]+]] !kernel_arg_access_qual [[META19:![0-9]+]] !kernel_arg_type [[META20:![0-9]+]] !kernel_arg_base_type [[META20]] !kernel_arg_type_qual [[META21:![0-9]+]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT3X3_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT4X4_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT32X32_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT64X64_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT3X3]], ptr addrspace(5) [[MAT3X3_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT4X4]], ptr addrspace(5) [[MAT4X4_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT32X32]], ptr addrspace(5) [[MAT32X32_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT64X64]], ptr addrspace(5) [[MAT64X64_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT3X3_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP2:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT4X4_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP3:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT32X32_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP4:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT64X64_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_caller_kern(ptr addrspace(1) noundef align 4 [[TMP0]], ptr addrspace(1) noundef align 4 [[TMP1]], ptr addrspace(1) noundef align 4 [[TMP2]], ptr addrspace(1) noundef align 4 [[TMP3]], ptr addrspace(1) noundef align 4 [[TMP4]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_caller_kern(
// AMDGCN-SAME: ptr addrspace(1) noundef align 4 [[A:%.*]], ptr addrspace(1) noundef align 4 [[MAT3X3:%.*]], ptr addrspace(1) noundef align 4 [[MAT4X4:%.*]], ptr addrspace(1) noundef align 4 [[MAT32X32:%.*]], ptr addrspace(1) noundef align 4 [[MAT64X64:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META18]] !kernel_arg_access_qual [[META19]] !kernel_arg_type [[META20]] !kernel_arg_base_type [[META20]] !kernel_arg_type_qual [[META21]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[A_ADDR_I:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[A_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT3X3_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT4X4_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT32X32_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[MAT64X64_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    store ptr addrspace(1) [[A]], ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT3X3]], ptr addrspace(5) [[MAT3X3_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT4X4]], ptr addrspace(5) [[MAT4X4_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT32X32]], ptr addrspace(5) [[MAT32X32_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[MAT64X64]], ptr addrspace(5) [[MAT64X64_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern(ptr addrspace(1) noundef align 4 [[TMP0]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_callee_kern(ptr addrspace(1) noundef align 4 [[TMP1]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP2:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_with_optnone_attribute(ptr addrspace(1) noundef align 4 [[TMP2]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP3:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[TMP3]], ptr addrspace(5) [[A_ADDR_I]], align 8
// AMDGCN-NEXT:    [[TMP4:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[A_ADDR_I]], align 8
// AMDGCN-NEXT:    store i32 1, ptr addrspace(1) [[TMP4]], align 4
// AMDGCN-NEXT:    [[TMP5:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT3X3_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP6:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT4X4_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat3X3(ptr addrspace(1) noundef align 4 [[TMP5]], ptr addrspace(1) noundef align 4 [[TMP6]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP7:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT32X32_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP8:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT64X64_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_callee_kern_Mat32X32(ptr addrspace(1) noundef align 4 [[TMP7]], ptr addrspace(1) noundef align 4 [[TMP8]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP9:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT3X3_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP10:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT4X4_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_callee_kern_Mat3X3(ptr addrspace(1) noundef align 4 [[TMP9]], ptr addrspace(1) noundef align 4 [[TMP10]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP11:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT32X32_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP12:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[MAT64X64_ADDR]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_callee_kern_Mat32X32(ptr addrspace(1) noundef align 4 [[TMP11]], ptr addrspace(1) noundef align 4 [[TMP12]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @caller_kern2(
// AMDGCN-SAME: <2 x i32> [[STRUCTONEMEM_COERCE:%.*]], ptr addrspace(1) noundef align 8 [[GLOBAL_STRUCTONEMEM:%.*]], ptr addrspace(4) noundef byref([[STRUCT_STRUCTTWOMEMBER:%.*]]) align 8 [[TMP0:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META22:![0-9]+]] !kernel_arg_access_qual [[META23:![0-9]+]] !kernel_arg_type [[META24:![0-9]+]] !kernel_arg_base_type [[META24]] !kernel_arg_type_qual [[META25:![0-9]+]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[STRUCTONEMEM:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[STRUCTTWOMEM:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[GLOBAL_STRUCTONEMEM_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[STRUCTONEMEM]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[STRUCTONEMEM_COERCE]], ptr addrspace(5) [[COERCE_DIVE]], align 8
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) align 8 [[STRUCTTWOMEM]], ptr addrspace(4) align 8 [[TMP0]], i64 16, i1 false)
// AMDGCN-NEXT:    store ptr addrspace(1) [[GLOBAL_STRUCTONEMEM]], ptr addrspace(5) [[GLOBAL_STRUCTONEMEM_ADDR]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = load ptr addrspace(1), ptr addrspace(5) [[GLOBAL_STRUCTONEMEM_ADDR]], align 8
// AMDGCN-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[STRUCTONEMEM]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr addrspace(5) [[COERCE_DIVE1]], align 8
// AMDGCN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP4:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP3]], align 8
// AMDGCN-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 1
// AMDGCN-NEXT:    [[TMP6:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP5]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_caller_kern2(<2 x i32> [[TMP2]], ptr addrspace(1) noundef align 8 [[TMP1]], <2 x i32> [[TMP4]], <2 x i32> [[TMP6]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_caller_kern2(
// AMDGCN-SAME: <2 x i32> [[STRUCTONEMEM_COERCE:%.*]], ptr addrspace(1) noundef align 8 [[GLOBAL_STRUCTONEMEM:%.*]], <2 x i32> [[STRUCTTWOMEM_COERCE0:%.*]], <2 x i32> [[STRUCTTWOMEM_COERCE1:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META22]] !kernel_arg_access_qual [[META23]] !kernel_arg_type [[META24]] !kernel_arg_base_type [[META24]] !kernel_arg_type_qual [[META25]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[STRUCTONEMEM:%.*]] = alloca [[STRUCT_STRUCTONEMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[STRUCTTWOMEM:%.*]] = alloca [[STRUCT_STRUCTTWOMEMBER:%.*]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[GLOBAL_STRUCTONEMEM_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// AMDGCN-NEXT:    [[COERCE_DIVE:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[STRUCTONEMEM]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[STRUCTONEMEM_COERCE]], ptr addrspace(5) [[COERCE_DIVE]], align 8
// AMDGCN-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 0
// AMDGCN-NEXT:    store <2 x i32> [[STRUCTTWOMEM_COERCE0]], ptr addrspace(5) [[TMP0]], align 8
// AMDGCN-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 1
// AMDGCN-NEXT:    store <2 x i32> [[STRUCTTWOMEM_COERCE1]], ptr addrspace(5) [[TMP1]], align 8
// AMDGCN-NEXT:    store ptr addrspace(1) [[GLOBAL_STRUCTONEMEM]], ptr addrspace(5) [[GLOBAL_STRUCTONEMEM_ADDR]], align 8
// AMDGCN-NEXT:    [[COERCE_DIVE1:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[STRUCTONEMEM]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP2:%.*]] = load <2 x i32>, ptr addrspace(5) [[COERCE_DIVE1]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelOneMember(<2 x i32> [[TMP2]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[COERCE_DIVE2:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTONEMEMBER]], ptr addrspace(5) [[STRUCTONEMEM]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP3:%.*]] = load <2 x i32>, ptr addrspace(5) [[COERCE_DIVE2]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelOneMember(<2 x i32> [[TMP3]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP5:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP4]], align 8
// AMDGCN-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 1
// AMDGCN-NEXT:    [[TMP7:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP6]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelTwoMember(<2 x i32> [[TMP5]], <2 x i32> [[TMP7]]) #[[ATTR7]]
// AMDGCN-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 0
// AMDGCN-NEXT:    [[TMP9:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP8]], align 8
// AMDGCN-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT_STRUCTTWOMEMBER]], ptr addrspace(5) [[STRUCTTWOMEM]], i32 0, i32 1
// AMDGCN-NEXT:    [[TMP11:%.*]] = load <2 x i32>, ptr addrspace(5) [[TMP10]], align 8
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelTwoMember(<2 x i32> [[TMP9]], <2 x i32> [[TMP11]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local amdgpu_kernel void @caller_kern3(
// AMDGCN-SAME: ptr addrspace(4) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[TMP0:%.*]], ptr addrspace(4) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[TMP1:%.*]]) #[[ATTR2]] !kernel_arg_addr_space [[META26:![0-9]+]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META27:![0-9]+]] !kernel_arg_base_type [[META27]] !kernel_arg_type_qual [[META11]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[LARGESTRUCTONEMEM:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[LARGESTRUCTTWOMEM:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) align 8 [[LARGESTRUCTONEMEM]], ptr addrspace(4) align 8 [[TMP0]], i64 800, i1 false)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p4.i64(ptr addrspace(5) align 8 [[LARGESTRUCTTWOMEM]], ptr addrspace(4) align 8 [[TMP1]], i64 480, i1 false)
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_caller_kern3(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER]]) align 8 [[LARGESTRUCTONEMEM]], ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 8 [[LARGESTRUCTTWOMEM]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//
// AMDGCN: Function Attrs: convergent noinline norecurse nounwind optnone
// AMDGCN-LABEL: define dso_local void @__clang_ocl_kern_imp_caller_kern3(
// AMDGCN-SAME: ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER:%.*]]) align 8 [[TMP0:%.*]], ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER:%.*]]) align 8 [[TMP1:%.*]]) #[[ATTR3]] !kernel_arg_addr_space [[META26]] !kernel_arg_access_qual [[META9]] !kernel_arg_type [[META27]] !kernel_arg_base_type [[META27]] !kernel_arg_type_qual [[META11]] {
// AMDGCN-NEXT:  entry:
// AMDGCN-NEXT:    [[LARGESTRUCTONEMEM:%.*]] = alloca [[STRUCT_LARGESTRUCTONEMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    [[LARGESTRUCTTWOMEM:%.*]] = alloca [[STRUCT_LARGESTRUCTTWOMEMBER]], align 8, addrspace(5)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 8 [[LARGESTRUCTONEMEM]], ptr addrspace(5) align 8 [[TMP0]], i64 800, i1 false)
// AMDGCN-NEXT:    call void @llvm.memcpy.p5.p5.i64(ptr addrspace(5) align 8 [[LARGESTRUCTTWOMEM]], ptr addrspace(5) align 8 [[TMP1]], i64 480, i1 false)
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeOneMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER]]) align 8 [[LARGESTRUCTONEMEM]]) #[[ATTR7]]
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_KernelLargeTwoMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 8 [[LARGESTRUCTTWOMEM]]) #[[ATTR7]]
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelLargeOneMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTONEMEMBER]]) align 8 [[LARGESTRUCTONEMEM]]) #[[ATTR7]]
// AMDGCN-NEXT:    call void @__clang_ocl_kern_imp_ext_KernelLargeTwoMember(ptr addrspace(5) noundef byref([[STRUCT_LARGESTRUCTTWOMEMBER]]) align 8 [[LARGESTRUCTTWOMEM]]) #[[ATTR7]]
// AMDGCN-NEXT:    ret void
//
//.
// X86: [[META4]] = !{i32 1}
// X86: [[META5]] = !{!"none"}
// X86: [[META6]] = !{!"int*"}
// X86: [[META7]] = !{!""}
// X86: [[META8]] = !{i32 1, i32 1}
// X86: [[META9]] = !{!"none", !"none"}
// X86: [[META10]] = !{!"Mat3X3*", !"Mat4X4*"}
// X86: [[META11]] = !{!"", !""}
// X86: [[META12]] = !{!"Mat32X32*", !"Mat64X64*"}
// X86: [[META13]] = !{i32 0}
// X86: [[META14]] = !{!"struct StructOneMember"}
// X86: [[META15]] = !{!"struct LargeStructOneMember"}
// X86: [[META16]] = !{!"struct StructTwoMember"}
// X86: [[META17]] = !{!"struct LargeStructTwoMember"}
// X86: [[META18]] = !{i32 1, i32 1, i32 1, i32 1, i32 1}
// X86: [[META19]] = !{!"none", !"none", !"none", !"none", !"none"}
// X86: [[META20]] = !{!"int*", !"Mat3X3*", !"Mat4X4*", !"Mat32X32*", !"Mat64X64*"}
// X86: [[META21]] = !{!"", !"", !"", !"", !""}
// X86: [[META22]] = !{i32 0, i32 1, i32 0}
// X86: [[META23]] = !{!"none", !"none", !"none"}
// X86: [[META24]] = !{!"struct StructOneMember", !"struct StructOneMember*", !"struct StructTwoMember"}
// X86: [[META25]] = !{!"", !"", !""}
// X86: [[META26]] = !{i32 0, i32 0}
// X86: [[META27]] = !{!"struct LargeStructOneMember", !"struct LargeStructTwoMember"}
//.
// AMDGCN: [[META4]] = !{i32 1}
// AMDGCN: [[META5]] = !{!"none"}
// AMDGCN: [[META6]] = !{!"int*"}
// AMDGCN: [[META7]] = !{!""}
// AMDGCN: [[META8]] = !{i32 1, i32 1}
// AMDGCN: [[META9]] = !{!"none", !"none"}
// AMDGCN: [[META10]] = !{!"Mat3X3*", !"Mat4X4*"}
// AMDGCN: [[META11]] = !{!"", !""}
// AMDGCN: [[META12]] = !{!"Mat32X32*", !"Mat64X64*"}
// AMDGCN: [[META13]] = !{i32 0}
// AMDGCN: [[META14]] = !{!"struct StructOneMember"}
// AMDGCN: [[META15]] = !{!"struct LargeStructOneMember"}
// AMDGCN: [[META16]] = !{!"struct StructTwoMember"}
// AMDGCN: [[META17]] = !{!"struct LargeStructTwoMember"}
// AMDGCN: [[META18]] = !{i32 1, i32 1, i32 1, i32 1, i32 1}
// AMDGCN: [[META19]] = !{!"none", !"none", !"none", !"none", !"none"}
// AMDGCN: [[META20]] = !{!"int*", !"Mat3X3*", !"Mat4X4*", !"Mat32X32*", !"Mat64X64*"}
// AMDGCN: [[META21]] = !{!"", !"", !"", !"", !""}
// AMDGCN: [[META22]] = !{i32 0, i32 1, i32 0}
// AMDGCN: [[META23]] = !{!"none", !"none", !"none"}
// AMDGCN: [[META24]] = !{!"struct StructOneMember", !"struct StructOneMember*", !"struct StructTwoMember"}
// AMDGCN: [[META25]] = !{!"", !"", !""}
// AMDGCN: [[META26]] = !{i32 0, i32 0}
// AMDGCN: [[META27]] = !{!"struct LargeStructOneMember", !"struct LargeStructTwoMember"}
//.