llvm/clang/test/OpenMP/parallel_copyin_codegen.cpp

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -fopenmp -fnoopenmp-use-tls -x c++ -std=c++11 -triple x86_64-linux -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -fnoopenmp-use-tls -x c++ -triple x86_64-linux -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -std=c++11 -DLAMBDA -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -fblocks -DBLOCKS -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK4
// RUN: %clang_cc1 -verify -fopenmp -fnoopenmp-use-tls -x c++ -std=c++11 -DARRAY -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5

// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -fopenmp-simd -fnoopenmp-use-tls -x c++ -std=c++11 -triple x86_64-linux -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -fnoopenmp-use-tls -x c++ -triple x86_64-linux -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -std=c++11 -DLAMBDA -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -fblocks -DBLOCKS -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -fnoopenmp-use-tls -x c++ -std=c++11 -DARRAY -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"

// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-linux -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-linux -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
// RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK14
// RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DARRAY -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK15
// RUN: %clang_cc1 -verify -fopenmp -x c++ -DNESTED -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK16

// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-linux -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-linux -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DARRAY -triple x86_64-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// expected-no-diagnostics
#if !defined(ARRAY) && !defined(NESTED)
#ifndef HEADER
#define HEADER

volatile int g __attribute__((aligned(128))) = 1212;
#pragma omp threadprivate(g)

template <class T>
struct S {
  T f;
  S(T a) : f(a + g) {}
  S() : f(g) {}
  S &operator=(const S &) { return *this; };
  operator T() { return T(); }
  ~S() {}
};



template <typename T>
T tmain() {
  S<T> test;
  test = S<T>();
  static T t_var __attribute__((aligned(128))) = 333;
  static T vec[] __attribute__((aligned(128))) = {3, 3};
  static S<T> s_arr[] __attribute__((aligned(128))) = {1, 2};
  static S<T> var __attribute__((aligned(128))) (3);
#pragma omp threadprivate(t_var, vec, s_arr, var)
#pragma omp parallel copyin(t_var, vec, s_arr, var)
  {
    vec[0] = t_var;
    s_arr[0] = var;
  }
#pragma omp parallel copyin(t_var)
  {}
  return T();
}

int main() {
#ifdef LAMBDA
  [&]() {


#pragma omp parallel copyin(g)
  {

    // threadprivate_g = g;


    g = 1;

    [&]() {
      g = 2;

    }();
  }
  }();
  return 0;
#elif defined(BLOCKS)

  ^{


#pragma omp parallel copyin(g)
  {

    // threadprivate_g = g;


    g = 1;


    ^{
      g = 2;

    }();
  }
  }();
  return 0;
#else
  S<float> test;
  test = S<float>();
  static int t_var = 1122;
  static int vec[] = {1, 2};
  static S<float> s_arr[] = {1, 2};
  static S<float> var(3);
#pragma omp threadprivate(t_var, vec, s_arr, var)
#pragma omp parallel copyin(t_var, vec, s_arr, var)
  {
    vec[0] = t_var;
    s_arr[0] = var;
  }
#pragma omp parallel copyin(t_var) default(none)
  ++t_var;
  return tmain<int>();
#endif
}





// threadprivate_t_var = t_var;



// threadprivate_vec = vec;


// threadprivate_s_arr = s_arr;


// threadprivate_var = var;






// threadprivate_t_var = t_var;







// threadprivate_t_var = t_var;



// threadprivate_vec = vec;


// threadprivate_s_arr = s_arr;


// threadprivate_var = var;






// threadprivate_t_var = t_var;





#endif
#elif defined(ARRAY)

struct St {
  int a, b;
  St() : a(0), b(0) {}
  St &operator=(const St &) { return *this; };
  ~St() {}
};

void array_func() {
  static int a[2];
  static St s[2];


#pragma omp threadprivate(a, s)
#pragma omp parallel copyin(a, s)
  ;
}
#elif defined(NESTED)
int t_init();
int t = t_init();
#pragma omp threadprivate(t)
void foo() {
#pragma omp parallel
#pragma omp parallel copyin(t)
  ++t;
}

#endif // NESTED

// CHECK1-LABEL: define {{[^@]+}}@main
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]], ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR2:[0-9]+]]
// CHECK1-NEXT:    [[TMP0:%.*]] = load atomic i8, ptr @_ZGVZ4mainE5s_arr acquire, align 8
// CHECK1-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK1-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT_CHECK:%.*]], label [[INIT_END:%.*]], !prof [[PROF3:![0-9]+]]
// CHECK1:       init.check:
// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__cxa_guard_acquire(ptr @_ZGVZ4mainE5s_arr) #[[ATTR2]]
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[INIT:%.*]], label [[INIT_END]]
// CHECK1:       init:
// CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
// CHECK1-NEXT:    call void @__kmpc_threadprivate_register(ptr @[[GLOB1]], ptr @_ZZ4mainE5s_arr, ptr @.__kmpc_global_ctor_., ptr null, ptr @.__kmpc_global_dtor_.)
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ4mainE5s_arr, float noundef 1.000000e+00)
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S]], ptr @_ZZ4mainE5s_arr, i64 1), float noundef 2.000000e+00)
// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
// CHECK1-NEXT:    call void @__cxa_guard_release(ptr @_ZGVZ4mainE5s_arr) #[[ATTR2]]
// CHECK1-NEXT:    br label [[INIT_END]]
// CHECK1:       init.end:
// CHECK1-NEXT:    [[TMP4:%.*]] = load atomic i8, ptr @_ZGVZ4mainE3var acquire, align 8
// CHECK1-NEXT:    [[GUARD_UNINITIALIZED1:%.*]] = icmp eq i8 [[TMP4]], 0
// CHECK1-NEXT:    br i1 [[GUARD_UNINITIALIZED1]], label [[INIT_CHECK2:%.*]], label [[INIT_END5:%.*]], !prof [[PROF3]]
// CHECK1:       init.check2:
// CHECK1-NEXT:    [[TMP5:%.*]] = call i32 @__cxa_guard_acquire(ptr @_ZGVZ4mainE3var) #[[ATTR2]]
// CHECK1-NEXT:    [[TOBOOL3:%.*]] = icmp ne i32 [[TMP5]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL3]], label [[INIT4:%.*]], label [[INIT_END5]]
// CHECK1:       init4:
// CHECK1-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    call void @__kmpc_threadprivate_register(ptr @[[GLOB1]], ptr @_ZZ4mainE3var, ptr @.__kmpc_global_ctor_..1, ptr null, ptr @.__kmpc_global_dtor_..2)
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ4mainE3var, float noundef 3.000000e+00)
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @_ZZ4mainE3var, ptr @__dso_handle) #[[ATTR2]]
// CHECK1-NEXT:    call void @__cxa_guard_release(ptr @_ZGVZ4mainE3var) #[[ATTR2]]
// CHECK1-NEXT:    br label [[INIT_END5]]
// CHECK1:       init.end5:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 0, ptr @main.omp_outlined)
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 0, ptr @main.omp_outlined.3)
// CHECK1-NEXT:    [[CALL6:%.*]] = call noundef i32 @_Z5tmainIiET_v()
// CHECK1-NEXT:    store i32 [[CALL6]], ptr [[RETVAL]], align 4
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR2]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK1-NEXT:    ret i32 [[TMP8]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEaSERKS0_
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret ptr [[THIS1]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_ctor_.
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3:[0-9]+]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]], float noundef 1.000000e+00)
// CHECK1-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[TMP1]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    ret ptr [[TMP2]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_dtor_.
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[TMP1]], i64 2
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[TMP1]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done1:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @_ZZ4mainE5s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @_ZZ4mainE5s_arr
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done1:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_ctor_..1
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]], float noundef 3.000000e+00)
// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    ret ptr [[TMP2]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_dtor_..2
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]]) #[[ATTR2]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@main.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE5t_var, i64 4, ptr @_ZZ4mainE5t_var.cache.)
// CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = icmp ne i64 ptrtoint (ptr @_ZZ4mainE5t_var to i64), [[TMP3]]
// CHECK1-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK1:       copyin.not.master:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr @_ZZ4mainE5t_var, align 4
// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[TMP2]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE3vec, i64 8, ptr @_ZZ4mainE3vec.cache.)
// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP6]], ptr align 4 @_ZZ4mainE3vec, i64 8, i1 false)
// CHECK1-NEXT:    [[TMP7:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE5s_arr, i64 8, ptr @_ZZ4mainE5s_arr.cache.)
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[TMP7]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE1:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ @_ZZ4mainE5s_arr, [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE1]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done1:
// CHECK1-NEXT:    [[TMP9:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE3var, i64 4, ptr @_ZZ4mainE3var.cache.)
// CHECK1-NEXT:    [[CALL2:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TMP9]], ptr noundef nonnull align 4 dereferenceable(4) @_ZZ4mainE3var)
// CHECK1-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK1:       copyin.not.master.end:
// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP10:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE5t_var, i64 4, ptr @_ZZ4mainE5t_var.cache.)
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE3vec, i64 8, ptr @_ZZ4mainE3vec.cache.)
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP12]], i64 0, i64 0
// CHECK1-NEXT:    store i32 [[TMP11]], ptr [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE3var, i64 4, ptr @_ZZ4mainE3var.cache.)
// CHECK1-NEXT:    [[TMP14:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE5s_arr, i64 8, ptr @_ZZ4mainE5s_arr.cache.)
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[TMP14]], i64 0, i64 0
// CHECK1-NEXT:    [[CALL4:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX3]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP13]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@main.omp_outlined.3
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE5t_var, i64 4, ptr @_ZZ4mainE5t_var.cache.)
// CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = icmp ne i64 ptrtoint (ptr @_ZZ4mainE5t_var to i64), [[TMP3]]
// CHECK1-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK1:       copyin.not.master:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr @_ZZ4mainE5t_var, align 4
// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[TMP2]], align 4
// CHECK1-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK1:       copyin.not.master.end:
// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP6:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ4mainE5t_var, i64 4, ptr @_ZZ4mainE5t_var.cache.)
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP7]], 1
// CHECK1-NEXT:    store i32 [[INC]], ptr [[TMP6]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK1-SAME: () #[[ATTR1]] comdat {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]], ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR2]]
// CHECK1-NEXT:    [[TMP0:%.*]] = load atomic i8, ptr @_ZGVZ5tmainIiET_vE5s_arr acquire, align 8
// CHECK1-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK1-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT_CHECK:%.*]], label [[INIT_END:%.*]], !prof [[PROF3]]
// CHECK1:       init.check:
// CHECK1-NEXT:    [[TMP1:%.*]] = call i32 @__cxa_guard_acquire(ptr @_ZGVZ5tmainIiET_vE5s_arr) #[[ATTR2]]
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[INIT:%.*]], label [[INIT_END]]
// CHECK1:       init:
// CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    call void @__kmpc_threadprivate_register(ptr @[[GLOB1]], ptr @_ZZ5tmainIiET_vE5s_arr, ptr @.__kmpc_global_ctor_..4, ptr null, ptr @.__kmpc_global_dtor_..5)
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ5tmainIiET_vE5s_arr, i32 noundef 1)
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S_0]], ptr @_ZZ5tmainIiET_vE5s_arr, i64 1), i32 noundef 2)
// CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor.6, ptr null, ptr @__dso_handle) #[[ATTR2]]
// CHECK1-NEXT:    call void @__cxa_guard_release(ptr @_ZGVZ5tmainIiET_vE5s_arr) #[[ATTR2]]
// CHECK1-NEXT:    br label [[INIT_END]]
// CHECK1:       init.end:
// CHECK1-NEXT:    [[TMP4:%.*]] = load atomic i8, ptr @_ZGVZ5tmainIiET_vE3var acquire, align 8
// CHECK1-NEXT:    [[GUARD_UNINITIALIZED1:%.*]] = icmp eq i8 [[TMP4]], 0
// CHECK1-NEXT:    br i1 [[GUARD_UNINITIALIZED1]], label [[INIT_CHECK2:%.*]], label [[INIT_END5:%.*]], !prof [[PROF3]]
// CHECK1:       init.check2:
// CHECK1-NEXT:    [[TMP5:%.*]] = call i32 @__cxa_guard_acquire(ptr @_ZGVZ5tmainIiET_vE3var) #[[ATTR2]]
// CHECK1-NEXT:    [[TOBOOL3:%.*]] = icmp ne i32 [[TMP5]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL3]], label [[INIT4:%.*]], label [[INIT_END5]]
// CHECK1:       init4:
// CHECK1-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    call void @__kmpc_threadprivate_register(ptr @[[GLOB1]], ptr @_ZZ5tmainIiET_vE3var, ptr @.__kmpc_global_ctor_..7, ptr null, ptr @.__kmpc_global_dtor_..8)
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ5tmainIiET_vE3var, i32 noundef 3)
// CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIiED1Ev, ptr @_ZZ5tmainIiET_vE3var, ptr @__dso_handle) #[[ATTR2]]
// CHECK1-NEXT:    call void @__cxa_guard_release(ptr @_ZGVZ5tmainIiET_vE3var) #[[ATTR2]]
// CHECK1-NEXT:    br label [[INIT_END5]]
// CHECK1:       init.end5:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 0, ptr @_Z5tmainIiET_v.omp_outlined)
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 0, ptr @_Z5tmainIiET_v.omp_outlined.9)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR2]]
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP1:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP0]], ptr @g, i64 4, ptr @g.cache.)
// CHECK1-NEXT:    [[TMP2:%.*]] = load volatile i32, ptr [[TMP1]], align 128
// CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP2]] to float
// CHECK1-NEXT:    store float [[CONV]], ptr [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP1:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP0]], ptr @g, i64 4, ptr @g.cache.)
// CHECK1-NEXT:    [[TMP3:%.*]] = load volatile i32, ptr [[TMP2]], align 128
// CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP3]] to float
// CHECK1-NEXT:    [[ADD:%.*]] = fadd float [[TMP1]], [[CONV]]
// CHECK1-NEXT:    store float [[ADD]], ptr [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEaSERKS0_
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret ptr [[THIS1]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_ctor_..4
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]], i32 noundef 1)
// CHECK1-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[TMP1]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef 2)
// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    ret ptr [[TMP2]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_dtor_..5
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[TMP1]], i64 2
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[TMP1]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done1:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@__cxx_global_array_dtor.6
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S_0:%.*]], ptr @_ZZ5tmainIiET_vE5s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @_ZZ5tmainIiET_vE5s_arr
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done1:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_ctor_..7
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]], i32 noundef 3)
// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    ret ptr [[TMP2]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@.__kmpc_global_dtor_..8
// CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR3]] section ".text.startup" {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TMP1]]) #[[ATTR2]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v.omp_outlined
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE5t_var, i64 4, ptr @_ZZ5tmainIiET_vE5t_var.cache.)
// CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = icmp ne i64 ptrtoint (ptr @_ZZ5tmainIiET_vE5t_var to i64), [[TMP3]]
// CHECK1-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK1:       copyin.not.master:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr @_ZZ5tmainIiET_vE5t_var, align 128
// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[TMP2]], align 128
// CHECK1-NEXT:    [[TMP6:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE3vec, i64 8, ptr @_ZZ5tmainIiET_vE3vec.cache.)
// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[TMP6]], ptr align 128 @_ZZ5tmainIiET_vE3vec, i64 8, i1 false)
// CHECK1-NEXT:    [[TMP7:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE5s_arr, i64 8, ptr @_ZZ5tmainIiET_vE5s_arr.cache.)
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[TMP7]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE1:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ @_ZZ5tmainIiET_vE5s_arr, [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE1]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done1:
// CHECK1-NEXT:    [[TMP9:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE3var, i64 4, ptr @_ZZ5tmainIiET_vE3var.cache.)
// CHECK1-NEXT:    [[CALL2:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TMP9]], ptr noundef nonnull align 4 dereferenceable(4) @_ZZ5tmainIiET_vE3var)
// CHECK1-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK1:       copyin.not.master.end:
// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP10:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE5t_var, i64 4, ptr @_ZZ5tmainIiET_vE5t_var.cache.)
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 128
// CHECK1-NEXT:    [[TMP12:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE3vec, i64 8, ptr @_ZZ5tmainIiET_vE3vec.cache.)
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP12]], i64 0, i64 0
// CHECK1-NEXT:    store i32 [[TMP11]], ptr [[ARRAYIDX]], align 128
// CHECK1-NEXT:    [[TMP13:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE3var, i64 4, ptr @_ZZ5tmainIiET_vE3var.cache.)
// CHECK1-NEXT:    [[TMP14:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE5s_arr, i64 8, ptr @_ZZ5tmainIiET_vE5s_arr.cache.)
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[TMP14]], i64 0, i64 0
// CHECK1-NEXT:    [[CALL4:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX3]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP13]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v.omp_outlined.9
// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ5tmainIiET_vE5t_var, i64 4, ptr @_ZZ5tmainIiET_vE5t_var.cache.)
// CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = icmp ne i64 ptrtoint (ptr @_ZZ5tmainIiET_vE5t_var to i64), [[TMP3]]
// CHECK1-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK1:       copyin.not.master:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr @_ZZ5tmainIiET_vE5t_var, align 128
// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[TMP2]], align 128
// CHECK1-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK1:       copyin.not.master.end:
// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2]], i32 [[TMP1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP1:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP0]], ptr @g, i64 4, ptr @g.cache.)
// CHECK1-NEXT:    [[TMP2:%.*]] = load volatile i32, ptr [[TMP1]], align 128
// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP0]], ptr @g, i64 4, ptr @g.cache.)
// CHECK1-NEXT:    [[TMP3:%.*]] = load volatile i32, ptr [[TMP2]], align 128
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP3]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK3-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK3-NEXT:    ret i32 0
//
//
// CHECK4-LABEL: define {{[^@]+}}@main
// CHECK4-SAME: () #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr getelementptr inbounds nuw ([[STRUCT___BLOCK_LITERAL_GENERIC:%.*]], ptr @__block_literal_global, i32 0, i32 3), align 8
// CHECK4-NEXT:    call void [[TMP0]](ptr noundef @__block_literal_global)
// CHECK4-NEXT:    ret i32 0
//
//
// CHECK4-LABEL: define {{[^@]+}}@__main_block_invoke
// CHECK4-SAME: (ptr noundef [[DOTBLOCK_DESCRIPTOR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTBLOCK_DESCRIPTOR_ADDR:%.*]] = alloca ptr, align 8
// CHECK4-NEXT:    [[BLOCK_ADDR:%.*]] = alloca ptr, align 8
// CHECK4-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[DOTBLOCK_DESCRIPTOR_ADDR]], align 8
// CHECK4-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[BLOCK_ADDR]], align 8
// CHECK4-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1:[0-9]+]], i32 0, ptr @__main_block_invoke.omp_outlined)
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@__main_block_invoke.omp_outlined
// CHECK4-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK4-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK4-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @g, i64 4, ptr @g.cache.)
// CHECK4-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK4-NEXT:    [[TMP4:%.*]] = icmp ne i64 ptrtoint (ptr @g to i64), [[TMP3]]
// CHECK4-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK4:       copyin.not.master:
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, ptr @g, align 128
// CHECK4-NEXT:    store volatile i32 [[TMP5]], ptr [[TMP2]], align 128
// CHECK4-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK4:       copyin.not.master.end:
// CHECK4-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP1]])
// CHECK4-NEXT:    [[TMP6:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @g, i64 4, ptr @g.cache.)
// CHECK4-NEXT:    store volatile i32 1, ptr [[TMP6]], align 128
// CHECK4-NEXT:    [[TMP7:%.*]] = load ptr, ptr getelementptr inbounds nuw ([[STRUCT___BLOCK_LITERAL_GENERIC:%.*]], ptr @__block_literal_global.2, i32 0, i32 3), align 8
// CHECK4-NEXT:    call void [[TMP7]](ptr noundef @__block_literal_global.2)
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@g_block_invoke
// CHECK4-SAME: (ptr noundef [[DOTBLOCK_DESCRIPTOR:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTBLOCK_DESCRIPTOR_ADDR:%.*]] = alloca ptr, align 8
// CHECK4-NEXT:    [[BLOCK_ADDR:%.*]] = alloca ptr, align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
// CHECK4-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[DOTBLOCK_DESCRIPTOR_ADDR]], align 8
// CHECK4-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[BLOCK_ADDR]], align 8
// CHECK4-NEXT:    [[TMP1:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP0]], ptr @g, i64 4, ptr @g.cache.)
// CHECK4-NEXT:    store volatile i32 2, ptr [[TMP1]], align 128
// CHECK4-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_Z10array_funcv
// CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[TMP0:%.*]] = load atomic i8, ptr @_ZGVZ10array_funcvE1s acquire, align 8
// CHECK5-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK5-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT_CHECK:%.*]], label [[INIT_END:%.*]], !prof [[PROF3:![0-9]+]]
// CHECK5:       init.check:
// CHECK5-NEXT:    [[TMP1:%.*]] = call i32 @__cxa_guard_acquire(ptr @_ZGVZ10array_funcvE1s) #[[ATTR1:[0-9]+]]
// CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
// CHECK5-NEXT:    br i1 [[TOBOOL]], label [[INIT:%.*]], label [[INIT_END]]
// CHECK5:       init:
// CHECK5-NEXT:    [[TMP2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
// CHECK5-NEXT:    call void @__kmpc_threadprivate_register(ptr @[[GLOB1]], ptr @_ZZ10array_funcvE1s, ptr @.__kmpc_global_ctor_., ptr null, ptr @.__kmpc_global_dtor_.)
// CHECK5-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK5:       arrayctor.loop:
// CHECK5-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ @_ZZ10array_funcvE1s, [[INIT]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK5-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[ARRAYCTOR_CUR]])
// CHECK5-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK5-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], getelementptr inbounds ([[STRUCT_ST]], ptr @_ZZ10array_funcvE1s, i64 2)
// CHECK5-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK5:       arrayctor.cont:
// CHECK5-NEXT:    [[TMP3:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR1]]
// CHECK5-NEXT:    call void @__cxa_guard_release(ptr @_ZGVZ10array_funcvE1s) #[[ATTR1]]
// CHECK5-NEXT:    br label [[INIT_END]]
// CHECK5:       init.end:
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 0, ptr @_Z10array_funcv.omp_outlined)
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.__kmpc_global_ctor_.
// CHECK5-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR2:[0-9]+]] section "__TEXT,__StaticInit,regular,pure_instructions" {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.St], ptr [[TMP1]], i32 0, i32 0
// CHECK5-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK5-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK5:       arrayctor.loop:
// CHECK5-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK5-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[ARRAYCTOR_CUR]])
// CHECK5-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_ST]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK5-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK5-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK5:       arrayctor.cont:
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    ret ptr [[TMP2]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2StC1Ev
// CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    call void @_ZN2StC2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@.__kmpc_global_dtor_.
// CHECK5-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR2]] section "__TEXT,__StaticInit,regular,pure_instructions" {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[TMP1]], i64 2
// CHECK5-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK5:       arraydestroy.body:
// CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_ST]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK5-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR1]]
// CHECK5-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[TMP1]]
// CHECK5-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK5:       arraydestroy.done1:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2StD1Ev
// CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    call void @_ZN2StD2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]]) #[[ATTR1]]
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
// CHECK5-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR2]] section "__TEXT,__StaticInit,regular,pure_instructions" {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK5:       arraydestroy.body:
// CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_ST:%.*]], ptr @_ZZ10array_funcvE1s, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_ST]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK5-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR1]]
// CHECK5-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @_ZZ10array_funcvE1s
// CHECK5-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK5:       arraydestroy.done1:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_Z10array_funcv.omp_outlined
// CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ10array_funcvE1a, i64 8, ptr @_ZZ10array_funcvE1a.cache.)
// CHECK5-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK5-NEXT:    [[TMP4:%.*]] = icmp ne i64 ptrtoint (ptr @_ZZ10array_funcvE1a to i64), [[TMP3]]
// CHECK5-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK5:       copyin.not.master:
// CHECK5-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP2]], ptr align 4 @_ZZ10array_funcvE1a, i64 8, i1 false)
// CHECK5-NEXT:    [[TMP5:%.*]] = call ptr @__kmpc_threadprivate_cached(ptr @[[GLOB1]], i32 [[TMP1]], ptr @_ZZ10array_funcvE1s, i64 16, ptr @_ZZ10array_funcvE1s.cache.)
// CHECK5-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.St], ptr [[TMP5]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK5-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE1:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK5:       omp.arraycpy.body:
// CHECK5-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ @_ZZ10array_funcvE1s, [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK5-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK5-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(8) ptr @_ZN2StaSERKS_(ptr noundef nonnull align 4 dereferenceable(8) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(8) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK5-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_ST]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK5-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_ST]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK5-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE1]], label [[OMP_ARRAYCPY_BODY]]
// CHECK5:       omp.arraycpy.done1:
// CHECK5-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK5:       copyin.not.master.end:
// CHECK5-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[TMP1]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2StaSERKS_
// CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[TMP0:%.*]]) #[[ATTR0]] align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    ret ptr [[THIS1]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2StC2Ev
// CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    store i32 0, ptr [[A]], align 4
// CHECK5-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
// CHECK5-NEXT:    store i32 0, ptr [[B]], align 4
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN2StD2Ev
// CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK11-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]], ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR3:[0-9]+]]
// CHECK11-NEXT:    [[TMP0:%.*]] = load i8, ptr @_ZGVZ4mainE5s_arr, align 1
// CHECK11-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK11-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT_CHECK:%.*]], label [[INIT_END:%.*]], !prof [[PROF3:![0-9]+]]
// CHECK11:       init.check:
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ4mainE5s_arr, float noundef 1.000000e+00)
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S]], ptr @_ZZ4mainE5s_arr, i64 1), float noundef 2.000000e+00)
// CHECK11-NEXT:    [[TMP1:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]]
// CHECK11-NEXT:    store i8 1, ptr @_ZGVZ4mainE5s_arr, align 1
// CHECK11-NEXT:    br label [[INIT_END]]
// CHECK11:       init.end:
// CHECK11-NEXT:    [[TMP2:%.*]] = load i8, ptr @_ZGVZ4mainE3var, align 1
// CHECK11-NEXT:    [[GUARD_UNINITIALIZED1:%.*]] = icmp eq i8 [[TMP2]], 0
// CHECK11-NEXT:    br i1 [[GUARD_UNINITIALIZED1]], label [[INIT_CHECK2:%.*]], label [[INIT_END3:%.*]], !prof [[PROF3]]
// CHECK11:       init.check2:
// CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ4mainE3var, float noundef 3.000000e+00)
// CHECK11-NEXT:    [[TMP3:%.*]] = call i32 @__cxa_thread_atexit(ptr @_ZN1SIfED1Ev, ptr @_ZZ4mainE3var, ptr @__dso_handle) #[[ATTR3]]
// CHECK11-NEXT:    store i8 1, ptr @_ZGVZ4mainE3var, align 1
// CHECK11-NEXT:    br label [[INIT_END3]]
// CHECK11:       init.end3:
// CHECK11-NEXT:    [[TMP4:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5t_var)
// CHECK11-NEXT:    [[TMP5:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE3vec)
// CHECK11-NEXT:    [[TMP6:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5s_arr)
// CHECK11-NEXT:    [[TMP7:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE3var)
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 4, ptr @main.omp_outlined, ptr [[TMP4]], ptr [[TMP5]], ptr [[TMP6]], ptr [[TMP7]])
// CHECK11-NEXT:    [[TMP8:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5t_var)
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @main.omp_outlined.1, ptr [[TMP8]])
// CHECK11-NEXT:    [[CALL4:%.*]] = call noundef i32 @_Z5tmainIiET_v()
// CHECK11-NEXT:    store i32 [[CALL4]], ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR3]]
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP9]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEaSERKS0_
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    ret ptr [[THIS1]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR3]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
// CHECK11-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR2:[0-9]+]] section ".text.startup" {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @_ZZ4mainE5s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR3]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @_ZZ4mainE5s_arr
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done1:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@main.omp_outlined
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP4:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5t_var)
// CHECK11-NEXT:    [[TMP5:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK11-NEXT:    [[TMP6:%.*]] = ptrtoint ptr [[TMP4]] to i64
// CHECK11-NEXT:    [[TMP7:%.*]] = icmp ne i64 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[TMP7]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK11:       copyin.not.master:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK11-NEXT:    store i32 [[TMP8]], ptr [[TMP4]], align 4
// CHECK11-NEXT:    [[TMP9:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE3vec)
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP9]], ptr align 4 [[TMP1]], i64 8, i1 false)
// CHECK11-NEXT:    [[TMP10:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5s_arr)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[TMP10]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP11]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE1:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP11]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE1]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done1:
// CHECK11-NEXT:    [[TMP12:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE3var)
// CHECK11-NEXT:    [[CALL2:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TMP12]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP3]])
// CHECK11-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK11:       copyin.not.master.end:
// CHECK11-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
// CHECK11-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1:[0-9]+]], i32 [[TMP14]])
// CHECK11-NEXT:    [[TMP15:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5t_var)
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4
// CHECK11-NEXT:    [[TMP17:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE3vec)
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP17]], i64 0, i64 0
// CHECK11-NEXT:    store i32 [[TMP16]], ptr [[ARRAYIDX]], align 4
// CHECK11-NEXT:    [[TMP18:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE3var)
// CHECK11-NEXT:    [[TMP19:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5s_arr)
// CHECK11-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[TMP19]], i64 0, i64 0
// CHECK11-NEXT:    [[CALL4:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIfEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX3]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP18]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@main.omp_outlined.1
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR4]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5t_var)
// CHECK11-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK11-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP1]] to i64
// CHECK11-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
// CHECK11-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK11:       copyin.not.master:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP1]], align 4
// CHECK11-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK11:       copyin.not.master.end:
// CHECK11-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
// CHECK11-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1]], i32 [[TMP7]])
// CHECK11-NEXT:    [[TMP8:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ4mainE5t_var)
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
// CHECK11-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    store i32 [[INC]], ptr [[TMP8]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
// CHECK11-SAME: () #[[ATTR1]] comdat {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK11-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]], ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]])
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR3]]
// CHECK11-NEXT:    [[TMP0:%.*]] = load i8, ptr @_ZGVZ5tmainIiET_vE5s_arr, align 8
// CHECK11-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK11-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT_CHECK:%.*]], label [[INIT_END:%.*]], !prof [[PROF3]]
// CHECK11:       init.check:
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ5tmainIiET_vE5s_arr, i32 noundef 1)
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S_0]], ptr @_ZZ5tmainIiET_vE5s_arr, i64 1), i32 noundef 2)
// CHECK11-NEXT:    [[TMP1:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor.2, ptr null, ptr @__dso_handle) #[[ATTR3]]
// CHECK11-NEXT:    store i8 1, ptr @_ZGVZ5tmainIiET_vE5s_arr, align 8
// CHECK11-NEXT:    br label [[INIT_END]]
// CHECK11:       init.end:
// CHECK11-NEXT:    [[TMP2:%.*]] = load i8, ptr @_ZGVZ5tmainIiET_vE3var, align 8
// CHECK11-NEXT:    [[GUARD_UNINITIALIZED1:%.*]] = icmp eq i8 [[TMP2]], 0
// CHECK11-NEXT:    br i1 [[GUARD_UNINITIALIZED1]], label [[INIT_CHECK2:%.*]], label [[INIT_END3:%.*]], !prof [[PROF3]]
// CHECK11:       init.check2:
// CHECK11-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) @_ZZ5tmainIiET_vE3var, i32 noundef 3)
// CHECK11-NEXT:    [[TMP3:%.*]] = call i32 @__cxa_thread_atexit(ptr @_ZN1SIiED1Ev, ptr @_ZZ5tmainIiET_vE3var, ptr @__dso_handle) #[[ATTR3]]
// CHECK11-NEXT:    store i8 1, ptr @_ZGVZ5tmainIiET_vE3var, align 8
// CHECK11-NEXT:    br label [[INIT_END3]]
// CHECK11:       init.end3:
// CHECK11-NEXT:    [[TMP4:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5t_var)
// CHECK11-NEXT:    [[TMP5:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE3vec)
// CHECK11-NEXT:    [[TMP6:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5s_arr)
// CHECK11-NEXT:    [[TMP7:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE3var)
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 4, ptr @_Z5tmainIiET_v.omp_outlined, ptr [[TMP4]], ptr [[TMP5]], ptr [[TMP6]], ptr [[TMP7]])
// CHECK11-NEXT:    [[TMP8:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5t_var)
// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @_Z5tmainIiET_v.omp_outlined.3, ptr [[TMP8]])
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR3]]
// CHECK11-NEXT:    ret i32 0
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK11-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr [[TMP0]], align 128
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
// CHECK11-NEXT:    store float [[CONV]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK11-NEXT:    [[TMP2:%.*]] = load volatile i32, ptr [[TMP1]], align 128
// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP2]] to float
// CHECK11-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
// CHECK11-NEXT:    store float [[ADD]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEaSERKS0_
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    ret ptr [[THIS1]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR3]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@__cxx_global_array_dtor.2
// CHECK11-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR2]] section ".text.startup" {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK11:       arraydestroy.body:
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S_0:%.*]], ptr @_ZZ5tmainIiET_vE5s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK11-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR3]]
// CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @_ZZ5tmainIiET_vE5s_arr
// CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK11:       arraydestroy.done1:
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v.omp_outlined
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
// CHECK11-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
// CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP4:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5t_var)
// CHECK11-NEXT:    [[TMP5:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK11-NEXT:    [[TMP6:%.*]] = ptrtoint ptr [[TMP4]] to i64
// CHECK11-NEXT:    [[TMP7:%.*]] = icmp ne i64 [[TMP5]], [[TMP6]]
// CHECK11-NEXT:    br i1 [[TMP7]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK11:       copyin.not.master:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP0]], align 128
// CHECK11-NEXT:    store i32 [[TMP8]], ptr [[TMP4]], align 128
// CHECK11-NEXT:    [[TMP9:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE3vec)
// CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[TMP9]], ptr align 128 [[TMP1]], i64 8, i1 false)
// CHECK11-NEXT:    [[TMP10:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5s_arr)
// CHECK11-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[TMP10]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK11-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP11]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE1:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK11:       omp.arraycpy.body:
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK11-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK11-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP11]]
// CHECK11-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE1]], label [[OMP_ARRAYCPY_BODY]]
// CHECK11:       omp.arraycpy.done1:
// CHECK11-NEXT:    [[TMP12:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE3var)
// CHECK11-NEXT:    [[CALL2:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[TMP12]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP3]])
// CHECK11-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK11:       copyin.not.master.end:
// CHECK11-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
// CHECK11-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1]], i32 [[TMP14]])
// CHECK11-NEXT:    [[TMP15:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5t_var)
// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 128
// CHECK11-NEXT:    [[TMP17:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE3vec)
// CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP17]], i64 0, i64 0
// CHECK11-NEXT:    store i32 [[TMP16]], ptr [[ARRAYIDX]], align 128
// CHECK11-NEXT:    [[TMP18:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE3var)
// CHECK11-NEXT:    [[TMP19:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5s_arr)
// CHECK11-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[TMP19]], i64 0, i64 0
// CHECK11-NEXT:    [[CALL4:%.*]] = call noundef nonnull align 4 dereferenceable(4) ptr @_ZN1SIiEaSERKS0_(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYIDX3]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP18]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiET_v.omp_outlined.3
// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR4]] {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK11-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
// CHECK11-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @_ZZ5tmainIiET_vE5t_var)
// CHECK11-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK11-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP1]] to i64
// CHECK11-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
// CHECK11-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK11:       copyin.not.master:
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 128
// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP1]], align 128
// CHECK11-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK11:       copyin.not.master.end:
// CHECK11-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
// CHECK11-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1]], i32 [[TMP7]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK11-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr [[TMP0]], align 128
// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
// CHECK11-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK11-NEXT:    [[TMP2:%.*]] = load volatile i32, ptr [[TMP1]], align 128
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP2]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZTW1g
// CHECK11-SAME: () #[[ATTR8:[0-9]+]] comdat {
// CHECK11-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK11-NEXT:    ret ptr [[TMP1]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@main
// CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
// CHECK13-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK13-NEXT:    ret i32 0
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZTW1g
// CHECK13-SAME: () #[[ATTR6:[0-9]+]] comdat {
// CHECK13-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK13-NEXT:    ret ptr [[TMP1]]
//
//
// CHECK14-LABEL: define {{[^@]+}}@main
// CHECK14-SAME: () #[[ATTR1:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK14-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK14-NEXT:    [[TMP0:%.*]] = load ptr, ptr getelementptr inbounds nuw ([[STRUCT___BLOCK_LITERAL_GENERIC:%.*]], ptr @__block_literal_global, i32 0, i32 3), align 8
// CHECK14-NEXT:    call void [[TMP0]](ptr noundef @__block_literal_global)
// CHECK14-NEXT:    ret i32 0
//
//
// CHECK14-LABEL: define {{[^@]+}}@__main_block_invoke
// CHECK14-SAME: (ptr noundef [[DOTBLOCK_DESCRIPTOR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTBLOCK_DESCRIPTOR_ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    [[BLOCK_ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[DOTBLOCK_DESCRIPTOR_ADDR]], align 8
// CHECK14-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[BLOCK_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK14-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 1, ptr @__main_block_invoke.omp_outlined, ptr [[TMP0]])
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@__main_block_invoke.omp_outlined
// CHECK14-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[G:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK14-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[G_ADDR]], align 8
// CHECK14-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK14-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK14-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP1]] to i64
// CHECK14-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
// CHECK14-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK14:       copyin.not.master:
// CHECK14-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 128
// CHECK14-NEXT:    store volatile i32 [[TMP5]], ptr [[TMP1]], align 128
// CHECK14-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK14:       copyin.not.master.end:
// CHECK14-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK14-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
// CHECK14-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1:[0-9]+]], i32 [[TMP7]])
// CHECK14-NEXT:    [[TMP8:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK14-NEXT:    store volatile i32 1, ptr [[TMP8]], align 128
// CHECK14-NEXT:    [[TMP9:%.*]] = load ptr, ptr getelementptr inbounds nuw ([[STRUCT___BLOCK_LITERAL_GENERIC:%.*]], ptr @__block_literal_global.2, i32 0, i32 3), align 8
// CHECK14-NEXT:    call void [[TMP9]](ptr noundef @__block_literal_global.2)
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@g_block_invoke
// CHECK14-SAME: (ptr noundef [[DOTBLOCK_DESCRIPTOR:%.*]]) #[[ATTR2]] {
// CHECK14-NEXT:  entry:
// CHECK14-NEXT:    [[DOTBLOCK_DESCRIPTOR_ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    [[BLOCK_ADDR:%.*]] = alloca ptr, align 8
// CHECK14-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[DOTBLOCK_DESCRIPTOR_ADDR]], align 8
// CHECK14-NEXT:    store ptr [[DOTBLOCK_DESCRIPTOR]], ptr [[BLOCK_ADDR]], align 8
// CHECK14-NEXT:    [[TMP0:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK14-NEXT:    store volatile i32 2, ptr [[TMP0]], align 128
// CHECK14-NEXT:    ret void
//
//
// CHECK14-LABEL: define {{[^@]+}}@_ZTW1g
// CHECK14-SAME: () #[[ATTR7:[0-9]+]] comdat {
// CHECK14-NEXT:    [[TMP1:%.*]] = call align 128 ptr @llvm.threadlocal.address.p0(ptr align 128 @g)
// CHECK14-NEXT:    ret ptr [[TMP1]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@_Z10array_funcv
// CHECK15-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[TMP0:%.*]] = load i8, ptr @_ZGVZ10array_funcvE1s, align 1
// CHECK15-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK15-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT_CHECK:%.*]], label [[INIT_END:%.*]], !prof [[PROF3:![0-9]+]]
// CHECK15:       init.check:
// CHECK15-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK15:       arrayctor.loop:
// CHECK15-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi ptr [ @_ZZ10array_funcvE1s, [[INIT_CHECK]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK15-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[ARRAYCTOR_CUR]])
// CHECK15-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[ARRAYCTOR_CUR]], i64 1
// CHECK15-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq ptr [[ARRAYCTOR_NEXT]], getelementptr inbounds ([[STRUCT_ST]], ptr @_ZZ10array_funcvE1s, i64 2)
// CHECK15-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK15:       arrayctor.cont:
// CHECK15-NEXT:    [[TMP1:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
// CHECK15-NEXT:    store i8 1, ptr @_ZGVZ10array_funcvE1s, align 1
// CHECK15-NEXT:    br label [[INIT_END]]
// CHECK15:       init.end:
// CHECK15-NEXT:    [[TMP2:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ10array_funcvE1a)
// CHECK15-NEXT:    [[TMP3:%.*]] = call align 16 ptr @llvm.threadlocal.address.p0(ptr align 16 @_ZZ10array_funcvE1s)
// CHECK15-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 2, ptr @_Z10array_funcv.omp_outlined, ptr [[TMP2]], ptr [[TMP3]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN2StC1Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    call void @_ZN2StC2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
// CHECK15-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR1:[0-9]+]] section ".text.startup" {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK15-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK15:       arraydestroy.body:
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_ST:%.*]], ptr @_ZZ10array_funcvE1s, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK15-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_ST]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK15-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
// CHECK15-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @_ZZ10array_funcvE1s
// CHECK15-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK15:       arraydestroy.done1:
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN2StD1Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    call void @_ZN2StD2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]]) #[[ATTR2]]
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_Z10array_funcv.omp_outlined
// CHECK15-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[A:%.*]], ptr noundef nonnull align 4 dereferenceable(16) [[S:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK15-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK15-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
// CHECK15-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK15-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ADDR]], align 8
// CHECK15-NEXT:    [[TMP2:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @_ZZ10array_funcvE1a)
// CHECK15-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK15-NEXT:    [[TMP4:%.*]] = ptrtoint ptr [[TMP2]] to i64
// CHECK15-NEXT:    [[TMP5:%.*]] = icmp ne i64 [[TMP3]], [[TMP4]]
// CHECK15-NEXT:    br i1 [[TMP5]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK15:       copyin.not.master:
// CHECK15-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP2]], ptr align 4 [[TMP0]], i64 8, i1 false)
// CHECK15-NEXT:    [[TMP6:%.*]] = call align 16 ptr @llvm.threadlocal.address.p0(ptr align 16 @_ZZ10array_funcvE1s)
// CHECK15-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.St], ptr [[TMP6]], i32 0, i32 0
// CHECK15-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[ARRAY_BEGIN]], i64 2
// CHECK15-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE1:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK15:       omp.arraycpy.body:
// CHECK15-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK15-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[COPYIN_NOT_MASTER]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK15-NEXT:    [[CALL:%.*]] = call noundef nonnull align 4 dereferenceable(8) ptr @_ZN2StaSERKS_(ptr noundef nonnull align 4 dereferenceable(8) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(8) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK15-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_ST]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK15-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_ST]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK15-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP7]]
// CHECK15-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE1]], label [[OMP_ARRAYCPY_BODY]]
// CHECK15:       omp.arraycpy.done1:
// CHECK15-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK15:       copyin.not.master.end:
// CHECK15-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK15-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
// CHECK15-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1:[0-9]+]], i32 [[TMP9]])
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN2StaSERKS_
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[TMP0:%.*]]) #[[ATTR0]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    ret ptr [[THIS1]]
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN2StC2Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK15-NEXT:    store i32 0, ptr [[A]], align 4
// CHECK15-NEXT:    [[B:%.*]] = getelementptr inbounds nuw [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
// CHECK15-NEXT:    store i32 0, ptr [[B]], align 4
// CHECK15-NEXT:    ret void
//
//
// CHECK15-LABEL: define {{[^@]+}}@_ZN2StD2Ev
// CHECK15-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR0]] comdat align 2 {
// CHECK15-NEXT:  entry:
// CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK15-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK15-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@__cxx_global_var_init
// CHECK16-SAME: () #[[ATTR0:[0-9]+]] section ".text.startup" {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z6t_initv()
// CHECK16-NEXT:    [[TMP0:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @t)
// CHECK16-NEXT:    store i32 [[CALL]], ptr [[TMP0]], align 4
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@_Z3foov
// CHECK16-SAME: () #[[ATTR3:[0-9]+]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 0, ptr @_Z3foov.omp_outlined)
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined
// CHECK16-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK16-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = call ptr @_ZTW1t()
// CHECK16-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @_Z3foov.omp_outlined.omp_outlined, ptr [[TMP0]])
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined
// CHECK16-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T:%.*]]) #[[ATTR4]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK16-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK16-NEXT:    [[T_ADDR:%.*]] = alloca ptr, align 8
// CHECK16-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK16-NEXT:    store ptr [[T]], ptr [[T_ADDR]], align 8
// CHECK16-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_ADDR]], align 8
// CHECK16-NEXT:    [[TMP1:%.*]] = call ptr @_ZTW1t()
// CHECK16-NEXT:    [[TMP2:%.*]] = ptrtoint ptr [[TMP0]] to i64
// CHECK16-NEXT:    [[TMP3:%.*]] = ptrtoint ptr [[TMP1]] to i64
// CHECK16-NEXT:    [[TMP4:%.*]] = icmp ne i64 [[TMP2]], [[TMP3]]
// CHECK16-NEXT:    br i1 [[TMP4]], label [[COPYIN_NOT_MASTER:%.*]], label [[COPYIN_NOT_MASTER_END:%.*]]
// CHECK16:       copyin.not.master:
// CHECK16-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK16-NEXT:    store i32 [[TMP5]], ptr [[TMP1]], align 4
// CHECK16-NEXT:    br label [[COPYIN_NOT_MASTER_END]]
// CHECK16:       copyin.not.master.end:
// CHECK16-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK16-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
// CHECK16-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB1:[0-9]+]], i32 [[TMP7]])
// CHECK16-NEXT:    [[TMP8:%.*]] = call ptr @_ZTW1t()
// CHECK16-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
// CHECK16-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK16-NEXT:    store i32 [[INC]], ptr [[TMP8]], align 4
// CHECK16-NEXT:    ret void
//
//
// CHECK16-LABEL: define {{[^@]+}}@_ZTW1t
// CHECK16-SAME: () #[[ATTR5:[0-9]+]] comdat {
// CHECK16-NEXT:    call void @_ZTH1t()
// CHECK16-NEXT:    [[TMP1:%.*]] = call align 4 ptr @llvm.threadlocal.address.p0(ptr align 4 @t)
// CHECK16-NEXT:    ret ptr [[TMP1]]
//
//
// CHECK16-LABEL: define {{[^@]+}}@__tls_init
// CHECK16-SAME: () #[[ATTR0]] {
// CHECK16-NEXT:  entry:
// CHECK16-NEXT:    [[TMP0:%.*]] = load i8, ptr @__tls_guard, align 1
// CHECK16-NEXT:    [[GUARD_UNINITIALIZED:%.*]] = icmp eq i8 [[TMP0]], 0
// CHECK16-NEXT:    br i1 [[GUARD_UNINITIALIZED]], label [[INIT:%.*]], label [[EXIT:%.*]], !prof [[PROF5:![0-9]+]]
// CHECK16:       init:
// CHECK16-NEXT:    store i8 1, ptr @__tls_guard, align 1
// CHECK16-NEXT:    call void @__cxx_global_var_init()
// CHECK16-NEXT:    br label [[EXIT]]
// CHECK16:       exit:
// CHECK16-NEXT:    ret void
//