xref: /dports/devel/wasi-compiler-rt13/llvm-project-13.0.1.src/clang/test/OpenMP/for_reduction_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 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap %s --check-prefix=CHECK1
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap %s --check-prefix=CHECK2
// RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap %s --check-prefix=CHECK3
// RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap %s --check-prefix=CHECK4

// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -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-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// expected-no-diagnostics
#ifndef HEADER
#define HEADER

volatile double g, g_orig;
volatile double &g1 = g_orig;

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


template <typename T, int length>
T tmain() {
  T t;
  S<T> test;
  T t_var = T(), t_var1;
  T vec[] = {1, 2};
  S<T> s_arr[] = {1, 2};
  S<T> &var = test;
  S<T> var1;
  S<T> arr[length];
#pragma omp parallel
#pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
#pragma omp parallel
#pragma omp for reduction(&& : t_var)
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
#pragma omp parallel
#pragma omp for reduction(+ : arr[1:length-2])
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
  return T();
}

extern S<float> **foo();

int main() {
#ifdef LAMBDA
  [&]() {
#pragma omp parallel
#pragma omp for reduction(+:g, g1)
    for (int i = 0; i < 2; ++i) {

    // Reduction list for runtime.

    g = 1;
    g1 = 1;

    [&]() {
      g = 2;
      g1 = 2;
    }();
  }
  }();
  return 0;
#elif defined(BLOCKS)
  ^{
#pragma omp parallel
#pragma omp for reduction(-:g, g1)
    for (int i = 0; i < 2; ++i)  {

    // Reduction list for runtime.

    g = 1;
    g1 = 1;

    ^{
      g = 2;
      g1 = 2;
    }();
  }
  }();
  return 0;
#else
  S<float> test;
  float t_var = 0, t_var1;
  int vec[] = {1, 2};
  S<float> s_arr[] = {1, 2, 3, 4};
  S<float> &var = test;
  S<float> var1, arrs[10][4];
  S<float> **var2 = foo();
  S<float> vvar2[5];
  S<float> (&var3)[4] = s_arr;
#pragma omp parallel
#pragma omp for reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1)
  for (int i = 0; i < 2; ++i) {
    vec[i] = t_var;
    s_arr[i] = var;
  }
  int arr[10][vec[1]];
#pragma omp parallel for reduction(+:arr[1][:vec[1]]) reduction(&:arrs[1:vec[1]][1:2])
  for (int i = 0; i < 10; ++i)
    ++arr[1][i];
#pragma omp parallel
#pragma omp for reduction(+:arr) reduction(&:arrs)
  for (int i = 0; i < 10; ++i)
    ++arr[1][i];
  // arr is a VLA, but the array section has constant length so we can generate a constant sized array!
#pragma omp parallel
#pragma omp for reduction(+:arr[1][0:2])
  for (int i = 0; i < 10; ++i)
    ++arr[1][i];
#pragma omp parallel
#pragma omp for reduction(& : var2[0 : 5][1 : 6])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : var2[1][1 : 6])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : var2[1 : 1][1 : 6])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : var2[1 : 1][1])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : vvar2[0 : 5])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : var3[1 : 2])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : var3[ : 2])
  for (int i = 0; i < 10; ++i)
    ;
  // TODO: The compiler should also be able to generate a constant sized array in this case!
#pragma omp parallel
#pragma omp for reduction(& : var3[2 : ])
  for (int i = 0; i < 10; ++i)
    ;
#pragma omp parallel
#pragma omp for reduction(& : var3)
  for (int i = 0; i < 10; ++i)
    ;
  return tmain<int, 42>();
#endif
}


// Reduction list for runtime.



// For + reduction operation initial value of private variable is 0.


// For & reduction operation initial value of private variable is ones in all bits.

// For && reduction operation initial value of private variable is 1.0.

// For min reduction operation initial value of private variable is largest repesentable value.


// Skip checks for internal operations.

// void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};


// res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);


// switch(res)

// case 1:
// t_var += t_var_reduction;

// var = var.operator &(var_reduction);

// var1 = var1.operator &&(var1_reduction);

// t_var1 = min(t_var1, t_var1_reduction);

// __kmpc_end_reduce(<loc>, <gtid>, &<lock>);

// break;

// case 2:
// t_var += t_var_reduction;

// var = var.operator &(var_reduction);

// var1 = var1.operator &&(var1_reduction);

// t_var1 = min(t_var1, t_var1_reduction);

// __kmpc_end_reduce(<loc>, <gtid>, &<lock>);

// break;


// void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
//  *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
//  ...
//  *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
//  *(Type<n>-1*)rhs[<n>-1]);
// }
// t_var_lhs = (float*)lhs[0];
// t_var_rhs = (float*)rhs[0];

// var_lhs = (S<float>*)lhs[1];
// var_rhs = (S<float>*)rhs[1];

// var1_lhs = (S<float>*)lhs[2];
// var1_rhs = (S<float>*)rhs[2];

// t_var1_lhs = (float*)lhs[3];
// t_var1_rhs = (float*)rhs[3];

// t_var_lhs += t_var_rhs;

// var_lhs = var_lhs.operator &(var_rhs);

// var1_lhs = var1_lhs.operator &&(var1_rhs);

// t_var1_lhs = min(t_var1_lhs, t_var1_rhs);


// Reduction list for runtime.



// Check initialization of private copy.


// Check initialization of private copy.

// Skip checks for internal operations.

// void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};


// res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);


// switch(res)

// case 1:

// arr[:] += arr_reduction[:];

// arrs[:] = var.operator &(arrs_reduction[:]);

// __kmpc_end_reduce(<loc>, <gtid>, &<lock>);

// break;

// case 2:

// arr[:] += arr_reduction[:];

// arrs[:] = var.operator &(arrs_reduction[:]);

// break;

// Check destruction of private copy.


// void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
//  *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
//  ...
//  *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
//  *(Type<n>-1*)rhs[<n>-1]);
// }
// arr_rhs = (int*)rhs[0];
// arr_lhs = (int*)lhs[0];

// arr_size = (size_t)lhs[1];

// arrs_rhs = (S<float>*)rhs[2];
// arrs_lhs = (S<float>*)lhs[2];

// arrs_size = (size_t)lhs[3];

// arr_lhs[:] += arr_rhs[:];

// arrs_lhs = arrs_lhs.operator &(arrs_rhs);




// Reduction list for runtime.



// Check initialization of private copy.

// Check initialization of private copy.

// Skip checks for internal operations.

// void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};


// res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);


// switch(res)

// case 1:

// arr[:] += arr_reduction[:];

// arrs[:] = var.operator &(arrs_reduction[:]);

// __kmpc_end_reduce(<loc>, <gtid>, &<lock>);

// break;

// case 2:

// arr[:] += arr_reduction[:];

// arrs[:] = var.operator &(arrs_reduction[:]);

// break;

// Check destruction of private copy.


// void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
//  *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
//  ...
//  *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
//  *(Type<n>-1*)rhs[<n>-1]);
// }
// arr_rhs = (int*)rhs[0];
// arr_lhs = (int*)lhs[0];

// arr_size = (size_t)lhs[1];

// arrs_rhs = (S<float>*)rhs[2];
// arrs_lhs = (S<float>*)lhs[2];

// arr_lhs[:] += arr_rhs[:];

// arrs_lhs = arrs_lhs.operator &(arrs_rhs);




// Reduction list for runtime.







// Reduction list for runtime.





// Reduction list for runtime.






// Reduction list for runtime.






// Reduction list for runtime.






// Reduction list for runtime.





// Reduction list for runtime.









// Reduction list for runtime.









// Reduction list for runtime.







// Reduction list for runtime.





// Not interested in this one:

// Reduction list for runtime.



// For + reduction operation initial value of private variable is 0.

// For & reduction operation initial value of private variable is ones in all bits.

// For && reduction operation initial value of private variable is 1.0.

// For min reduction operation initial value of private variable is largest repesentable value.

// Skip checks for internal operations.

// void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};


// res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);


// switch(res)

// case 1:
// t_var += t_var_reduction;

// var = var.operator &(var_reduction);

// var1 = var1.operator &&(var1_reduction);

// t_var1 = min(t_var1, t_var1_reduction);

// __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>);

// break;

// case 2:
// t_var += t_var_reduction;

// var = var.operator &(var_reduction);

// var1 = var1.operator &&(var1_reduction);

// t_var1 = min(t_var1, t_var1_reduction);

// break;

// void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
//  *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
//  ...
//  *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
//  *(Type<n>-1*)rhs[<n>-1]);
// }
// t_var_lhs = (i{{[0-9]+}}*)lhs[0];
// t_var_rhs = (i{{[0-9]+}}*)rhs[0];

// var_lhs = (S<i{{[0-9]+}}>*)lhs[1];
// var_rhs = (S<i{{[0-9]+}}>*)rhs[1];

// var1_lhs = (S<i{{[0-9]+}}>*)lhs[2];
// var1_rhs = (S<i{{[0-9]+}}>*)rhs[2];

// t_var1_lhs = (i{{[0-9]+}}*)lhs[3];
// t_var1_rhs = (i{{[0-9]+}}*)rhs[3];

// t_var_lhs += t_var_rhs;

// var_lhs = var_lhs.operator &(var_rhs);

// var1_lhs = var1_lhs.operator &&(var1_rhs);

// t_var1_lhs = min(t_var1_lhs, t_var1_rhs);



// Reduction list for runtime.




#endif
// 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:    [[T_VAR:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [4 x %struct.S], align 16
// CHECK1-NEXT:    [[VAR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[VAR1:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK1-NEXT:    [[ARRS:%.*]] = alloca [10 x [4 x %struct.S]], align 16
// CHECK1-NEXT:    [[VAR2:%.*]] = alloca %struct.S**, align 8
// CHECK1-NEXT:    [[VVAR2:%.*]] = alloca [5 x %struct.S], align 16
// CHECK1-NEXT:    [[VAR3:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK1-NEXT:    store float 0.000000e+00, float* [[T_VAR]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const.main.vec to i8*), i64 8, i1 false)
// CHECK1-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[S_ARR]], i64 0, i64 0
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float 1.000000e+00)
// CHECK1-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_BEGIN]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK1-NEXT:    [[ARRAYINIT_ELEMENT1:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_ELEMENT]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT1]], float 3.000000e+00)
// CHECK1-NEXT:    [[ARRAYINIT_ELEMENT2:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_ELEMENT1]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT2]], float 4.000000e+00)
// CHECK1-NEXT:    store %struct.S* [[TEST]], %struct.S** [[VAR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR1]])
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 40
// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK1:       arrayctor.loop:
// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYCTOR_CUR]], i64 1
// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq %struct.S* [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK1:       arrayctor.cont:
// CHECK1-NEXT:    [[CALL:%.*]] = call %struct.S** @_Z3foov()
// CHECK1-NEXT:    store %struct.S** [[CALL]], %struct.S*** [[VAR2]], align 8
// CHECK1-NEXT:    [[ARRAY_BEGIN3:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR2]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYCTOR_END4:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN3]], i64 5
// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP5:%.*]]
// CHECK1:       arrayctor.loop5:
// CHECK1-NEXT:    [[ARRAYCTOR_CUR6:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN3]], [[ARRAYCTOR_CONT]] ], [ [[ARRAYCTOR_NEXT7:%.*]], [[ARRAYCTOR_LOOP5]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR6]])
// CHECK1-NEXT:    [[ARRAYCTOR_NEXT7]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYCTOR_CUR6]], i64 1
// CHECK1-NEXT:    [[ARRAYCTOR_DONE8:%.*]] = icmp eq %struct.S* [[ARRAYCTOR_NEXT7]], [[ARRAYCTOR_END4]]
// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE8]], label [[ARRAYCTOR_CONT9:%.*]], label [[ARRAYCTOR_LOOP5]]
// CHECK1:       arrayctor.cont9:
// CHECK1-NEXT:    store [4 x %struct.S]* [[S_ARR]], [4 x %struct.S]** [[VAR3]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float*, %struct.S*, %struct.S*, float*, [2 x i32]*, [4 x %struct.S]*)* @.omp_outlined. to void (i32*, i32*, ...)*), float* [[T_VAR]], %struct.S* [[TMP1]], %struct.S* [[VAR1]], float* [[T_VAR1]], [2 x i32]* [[VEC]], [4 x %struct.S]* [[S_ARR]])
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP4]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = mul nuw i64 10, [[TMP3]]
// CHECK1-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 16
// CHECK1-NEXT:    store i64 [[TMP3]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, [2 x i32]*, [10 x [4 x %struct.S]]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 10, i64 [[TMP3]], i32* [[VLA]], [2 x i32]* [[VEC]], [10 x [4 x %struct.S]]* [[ARRS]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, [10 x [4 x %struct.S]]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 10, i64 [[TMP3]], i32* [[VLA]], [10 x [4 x %struct.S]]* [[ARRS]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 10, i64 [[TMP3]], i32* [[VLA]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..13 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [5 x %struct.S]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), [5 x %struct.S]* [[VVAR2]])
// CHECK1-NEXT:    [[TMP6:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP6]])
// CHECK1-NEXT:    [[TMP7:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP7]])
// CHECK1-NEXT:    [[TMP8:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP8]])
// CHECK1-NEXT:    [[TMP9:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP9]])
// CHECK1-NEXT:    [[CALL10:%.*]] = call i32 @_Z5tmainIiLi42EET_v()
// CHECK1-NEXT:    store i32 [[CALL10]], i32* [[RETVAL]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP10]])
// CHECK1-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR2]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN11]], i64 5
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ARRAYCTOR_CONT9]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5:[0-9]+]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done12:
// CHECK1-NEXT:    [[ARRAY_BEGIN13:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN13]], i64 40
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY14:%.*]]
// CHECK1:       arraydestroy.body14:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST15:%.*]] = phi %struct.S* [ [[TMP12]], [[ARRAYDESTROY_DONE12]] ], [ [[ARRAYDESTROY_ELEMENT16:%.*]], [[ARRAYDESTROY_BODY14]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT16]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST15]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT16]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE17:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT16]], [[ARRAY_BEGIN13]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE17]], label [[ARRAYDESTROY_DONE18:%.*]], label [[ARRAYDESTROY_BODY14]]
// CHECK1:       arraydestroy.done18:
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR1]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAY_BEGIN19:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN19]], i64 4
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY20:%.*]]
// CHECK1:       arraydestroy.body20:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST21:%.*]] = phi %struct.S* [ [[TMP13]], [[ARRAYDESTROY_DONE18]] ], [ [[ARRAYDESTROY_ELEMENT22:%.*]], [[ARRAYDESTROY_BODY20]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT22]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST21]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT22]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE23:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT22]], [[ARRAY_BEGIN19]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE23]], label [[ARRAYDESTROY_DONE24:%.*]], label [[ARRAYDESTROY_BODY20]]
// CHECK1:       arraydestroy.done24:
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR5]]
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK1-NEXT:    ret i32 [[TMP14]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK1-NEXT:    call void @_ZN1SIfEC2Ef(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR1:%.*]], float* nonnull align 4 dereferenceable(4) [[T_VAR1:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[S_ARR:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca float*, align 8
// CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[VAR1_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[T_VAR1_ADDR:%.*]] = alloca float*, align 8
// CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[T_VAR3:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[VAR16:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK1-NEXT:    [[T_VAR17:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [4 x i8*], align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[_TMP22:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[REF_TMP25:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK1-NEXT:    [[ATOMIC_TEMP35:%.*]] = alloca float, align 4
// CHECK1-NEXT:    [[_TMP36:%.*]] = alloca float, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store float* [[T_VAR]], float** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S* [[VAR1]], %struct.S** [[VAR1_ADDR]], align 8
// CHECK1-NEXT:    store float* [[T_VAR1]], float** [[T_VAR1_ADDR]], align 8
// CHECK1-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[S_ARR]], [4 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load float*, float** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR1_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load float*, float** [[T_VAR1_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP1]], %struct.S** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP6]], %struct.S** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    store float 0.000000e+00, float* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[_TMP1]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK1-NEXT:    store %struct.S* [[VAR4]], %struct.S** [[_TMP5]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK1-NEXT:    store float 0x47EFFFFFE0000000, float* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP10]], 1
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = fptosi float [[TMP16]] to i32
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP4]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load %struct.S*, %struct.S** [[_TMP5]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP19]] to i64
// CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP5]], i64 0, i64 [[IDXPROM9]]
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast %struct.S* [[ARRAYIDX10]] to i8*
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast %struct.S* [[TMP18]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP20]], i8* align 4 [[TMP21]], i64 4, i1 false)
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP24:%.*]] = bitcast float* [[T_VAR3]] to i8*
// CHECK1-NEXT:    store i8* [[TMP24]], i8** [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[VAR4]] to i8*
// CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[VAR16]] to i8*
// CHECK1-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP30:%.*]] = bitcast float* [[T_VAR17]] to i8*
// CHECK1-NEXT:    store i8* [[TMP30]], i8** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = bitcast [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP32:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 4, i64 32, i8* [[TMP31]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP32]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP33:%.*]] = load float, float* [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP34:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[ADD12:%.*]] = fadd float [[TMP33]], [[TMP34]]
// CHECK1-NEXT:    store float [[ADD12]], float* [[TMP0]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK1-NEXT:    [[TMP35:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK1-NEXT:    [[CALL13:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[CALL13]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK1:       land.rhs:
// CHECK1-NEXT:    [[CALL14:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK1-NEXT:    [[TOBOOL15:%.*]] = fcmp une float [[CALL14]], 0.000000e+00
// CHECK1-NEXT:    br label [[LAND_END]]
// CHECK1:       land.end:
// CHECK1-NEXT:    [[TMP37:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL15]], [[LAND_RHS]] ]
// CHECK1-NEXT:    [[CONV16:%.*]] = uitofp i1 [[TMP37]] to float
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]], float [[CONV16]])
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[TMP2]] to i8*
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast %struct.S* [[REF_TMP]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP38]], i8* align 4 [[TMP39]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK1-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP3]], align 4
// CHECK1-NEXT:    [[TMP41:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[CMP17:%.*]] = fcmp olt float [[TMP40]], [[TMP41]]
// CHECK1-NEXT:    br i1 [[CMP17]], label [[COND_TRUE18:%.*]], label [[COND_FALSE19:%.*]]
// CHECK1:       cond.true18:
// CHECK1-NEXT:    [[TMP42:%.*]] = load float, float* [[TMP3]], align 4
// CHECK1-NEXT:    br label [[COND_END20:%.*]]
// CHECK1:       cond.false19:
// CHECK1-NEXT:    [[TMP43:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK1-NEXT:    br label [[COND_END20]]
// CHECK1:       cond.end20:
// CHECK1-NEXT:    [[COND21:%.*]] = phi float [ [[TMP42]], [[COND_TRUE18]] ], [ [[TMP43]], [[COND_FALSE19]] ]
// CHECK1-NEXT:    store float [[COND21]], float* [[TMP3]], align 4
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP44:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP45:%.*]] = bitcast float* [[TMP0]] to i32*
// CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, i32* [[TMP45]] monotonic, align 4
// CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
// CHECK1:       atomic_cont:
// CHECK1-NEXT:    [[TMP46:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP54:%.*]], [[ATOMIC_CONT]] ]
// CHECK1-NEXT:    [[TMP47:%.*]] = bitcast float* [[ATOMIC_TEMP]] to i32*
// CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i32 [[TMP46]] to float
// CHECK1-NEXT:    store float [[TMP48]], float* [[_TMP22]], align 4
// CHECK1-NEXT:    [[TMP49:%.*]] = load float, float* [[_TMP22]], align 4
// CHECK1-NEXT:    [[TMP50:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[ADD23:%.*]] = fadd float [[TMP49]], [[TMP50]]
// CHECK1-NEXT:    store float [[ADD23]], float* [[ATOMIC_TEMP]], align 4
// CHECK1-NEXT:    [[TMP51:%.*]] = load i32, i32* [[TMP47]], align 4
// CHECK1-NEXT:    [[TMP52:%.*]] = bitcast float* [[TMP0]] to i32*
// CHECK1-NEXT:    [[TMP53:%.*]] = cmpxchg i32* [[TMP52]], i32 [[TMP46]], i32 [[TMP51]] monotonic monotonic, align 4
// CHECK1-NEXT:    [[TMP54]] = extractvalue { i32, i1 } [[TMP53]], 0
// CHECK1-NEXT:    [[TMP55:%.*]] = extractvalue { i32, i1 } [[TMP53]], 1
// CHECK1-NEXT:    br i1 [[TMP55]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
// CHECK1:       atomic_exit:
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL24:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK1-NEXT:    [[TMP56:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
// CHECK1-NEXT:    [[TMP57:%.*]] = bitcast %struct.S* [[CALL24]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP56]], i8* align 4 [[TMP57]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL26:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK1-NEXT:    [[TOBOOL27:%.*]] = fcmp une float [[CALL26]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[TOBOOL27]], label [[LAND_RHS28:%.*]], label [[LAND_END31:%.*]]
// CHECK1:       land.rhs28:
// CHECK1-NEXT:    [[CALL29:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK1-NEXT:    [[TOBOOL30:%.*]] = fcmp une float [[CALL29]], 0.000000e+00
// CHECK1-NEXT:    br label [[LAND_END31]]
// CHECK1:       land.end31:
// CHECK1-NEXT:    [[TMP58:%.*]] = phi i1 [ false, [[ATOMIC_EXIT]] ], [ [[TOBOOL30]], [[LAND_RHS28]] ]
// CHECK1-NEXT:    [[CONV32:%.*]] = uitofp i1 [[TMP58]] to float
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP25]], float [[CONV32]])
// CHECK1-NEXT:    [[TMP59:%.*]] = bitcast %struct.S* [[TMP2]] to i8*
// CHECK1-NEXT:    [[TMP60:%.*]] = bitcast %struct.S* [[REF_TMP25]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP59]], i8* align 4 [[TMP60]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP25]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[TMP61:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[TMP62:%.*]] = bitcast float* [[TMP3]] to i32*
// CHECK1-NEXT:    [[ATOMIC_LOAD33:%.*]] = load atomic i32, i32* [[TMP62]] monotonic, align 4
// CHECK1-NEXT:    br label [[ATOMIC_CONT34:%.*]]
// CHECK1:       atomic_cont34:
// CHECK1-NEXT:    [[TMP63:%.*]] = phi i32 [ [[ATOMIC_LOAD33]], [[LAND_END31]] ], [ [[TMP73:%.*]], [[COND_END40:%.*]] ]
// CHECK1-NEXT:    [[TMP64:%.*]] = bitcast float* [[ATOMIC_TEMP35]] to i32*
// CHECK1-NEXT:    [[TMP65:%.*]] = bitcast i32 [[TMP63]] to float
// CHECK1-NEXT:    store float [[TMP65]], float* [[_TMP36]], align 4
// CHECK1-NEXT:    [[TMP66:%.*]] = load float, float* [[_TMP36]], align 4
// CHECK1-NEXT:    [[TMP67:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[CMP37:%.*]] = fcmp olt float [[TMP66]], [[TMP67]]
// CHECK1-NEXT:    br i1 [[CMP37]], label [[COND_TRUE38:%.*]], label [[COND_FALSE39:%.*]]
// CHECK1:       cond.true38:
// CHECK1-NEXT:    [[TMP68:%.*]] = load float, float* [[_TMP36]], align 4
// CHECK1-NEXT:    br label [[COND_END40]]
// CHECK1:       cond.false39:
// CHECK1-NEXT:    [[TMP69:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK1-NEXT:    br label [[COND_END40]]
// CHECK1:       cond.end40:
// CHECK1-NEXT:    [[COND41:%.*]] = phi float [ [[TMP68]], [[COND_TRUE38]] ], [ [[TMP69]], [[COND_FALSE39]] ]
// CHECK1-NEXT:    store float [[COND41]], float* [[ATOMIC_TEMP35]], align 4
// CHECK1-NEXT:    [[TMP70:%.*]] = load i32, i32* [[TMP64]], align 4
// CHECK1-NEXT:    [[TMP71:%.*]] = bitcast float* [[TMP3]] to i32*
// CHECK1-NEXT:    [[TMP72:%.*]] = cmpxchg i32* [[TMP71]], i32 [[TMP63]], i32 [[TMP70]] monotonic monotonic, align 4
// CHECK1-NEXT:    [[TMP73]] = extractvalue { i32, i1 } [[TMP72]], 0
// CHECK1-NEXT:    [[TMP74:%.*]] = extractvalue { i32, i1 } [[TMP72]], 1
// CHECK1-NEXT:    br i1 [[TMP74]], label [[ATOMIC_EXIT42:%.*]], label [[ATOMIC_CONT34]]
// CHECK1:       atomic_exit42:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP9]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [4 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [4 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to float*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to float*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to %struct.S*
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S*
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S*
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP22]] to %struct.S*
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8* [[TMP25]] to float*
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP28:%.*]] = load i8*, i8** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8* [[TMP28]] to float*
// CHECK1-NEXT:    [[TMP30:%.*]] = load float, float* [[TMP11]], align 4
// CHECK1-NEXT:    [[TMP31:%.*]] = load float, float* [[TMP8]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = fadd float [[TMP30]], [[TMP31]]
// CHECK1-NEXT:    store float [[ADD]], float* [[TMP11]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP17]], %struct.S* nonnull align 4 dereferenceable(4) [[TMP14]])
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[TMP17]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK1-NEXT:    [[CALL2:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP23]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[CALL2]], 0.000000e+00
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK1:       land.rhs:
// CHECK1-NEXT:    [[CALL3:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP20]])
// CHECK1-NEXT:    [[TOBOOL4:%.*]] = fcmp une float [[CALL3]], 0.000000e+00
// CHECK1-NEXT:    br label [[LAND_END]]
// CHECK1:       land.end:
// CHECK1-NEXT:    [[TMP34:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL4]], [[LAND_RHS]] ]
// CHECK1-NEXT:    [[CONV:%.*]] = uitofp i1 [[TMP34]] to float
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]], float [[CONV]])
// CHECK1-NEXT:    [[TMP35:%.*]] = bitcast %struct.S* [[TMP23]] to i8*
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast %struct.S* [[REF_TMP]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK1-NEXT:    [[TMP37:%.*]] = load float, float* [[TMP29]], align 4
// CHECK1-NEXT:    [[TMP38:%.*]] = load float, float* [[TMP26]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = fcmp olt float [[TMP37]], [[TMP38]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP39:%.*]] = load float, float* [[TMP29]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP26]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi float [ [[TMP39]], [[COND_TRUE]] ], [ [[TMP40]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store float [[COND]], float* [[TMP29]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEanERKS0_
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR7:[0-9]+]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret %struct.S* [[THIS1]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEcvfEv
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) #[[ATTR7]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret float 0.000000e+00
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIfED2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR5]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [10 x [4 x %struct.S]]* nonnull align 4 dereferenceable(160) [[ARRS:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[ARR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK1-NEXT:    [[ARRS_ADDR:%.*]] = alloca [10 x [4 x %struct.S]]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [4 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i32* [[ARR]], i32** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    store [10 x [4 x %struct.S]]* [[ARRS]], [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load [10 x [4 x %struct.S]]*, [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP5]]
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 0
// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
// CHECK1-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
// CHECK1-NEXT:    [[TMP8:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP8]]
// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX5]], i64 [[LB_ADD_LEN]]
// CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint i32* [[ARRAYIDX6]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = ptrtoint i32* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    [[TMP12:%.*]] = sdiv exact i64 [[TMP11]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP13:%.*]] = add nuw i64 [[TMP12]], 1
// CHECK1-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP15:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP15]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[VLA7:%.*]] = alloca i32, i64 [[TMP13]], align 16
// CHECK1-NEXT:    store i64 [[TMP13]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr i32, i32* [[VLA7]], i64 [[TMP13]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[VLA7]], [[TMP16]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[VLA7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP16]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP17:%.*]] = ptrtoint i32* [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i32* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
// CHECK1-NEXT:    [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[VLA7]], i64 [[TMP20]]
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[TMP4]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[ARRAYIDX8]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[ARRAYDECAY]], i64 1
// CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[ARRAYIDX10]], align 4
// CHECK1-NEXT:    [[TMP23:%.*]] = sext i32 [[TMP22]] to i64
// CHECK1-NEXT:    [[LB_ADD_LEN11:%.*]] = add nsw i64 0, [[TMP23]]
// CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[TMP4]], i64 0, i64 [[LB_ADD_LEN11]]
// CHECK1-NEXT:    [[ARRAYDECAY13:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[ARRAYIDX12]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDECAY13]], i64 2
// CHECK1-NEXT:    [[TMP24:%.*]] = ptrtoint %struct.S* [[ARRAYIDX14]] to i64
// CHECK1-NEXT:    [[TMP25:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
// CHECK1-NEXT:    [[TMP26:%.*]] = sub i64 [[TMP24]], [[TMP25]]
// CHECK1-NEXT:    [[TMP27:%.*]] = sdiv exact i64 [[TMP26]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP28:%.*]] = add nuw i64 [[TMP27]], 1
// CHECK1-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[VLA15:%.*]] = alloca [[STRUCT_S]], i64 [[TMP28]], align 16
// CHECK1-NEXT:    store i64 [[TMP28]], i64* [[__VLA_EXPR1]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP28]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY16:%.*]] = icmp eq %struct.S* [[VLA15]], [[TMP30]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY16]], label [[OMP_ARRAYINIT_DONE21:%.*]], label [[OMP_ARRAYINIT_BODY17:%.*]]
// CHECK1:       omp.arrayinit.body17:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST18:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYINIT_DONE]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT19:%.*]], [[OMP_ARRAYINIT_BODY17]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST18]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST18]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT19]], [[TMP30]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYINIT_DONE21]], label [[OMP_ARRAYINIT_BODY17]]
// CHECK1:       omp.arrayinit.done21:
// CHECK1-NEXT:    [[TMP31:%.*]] = bitcast [10 x [4 x %struct.S]]* [[TMP4]] to %struct.S*
// CHECK1-NEXT:    [[TMP32:%.*]] = ptrtoint %struct.S* [[TMP31]] to i64
// CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
// CHECK1-NEXT:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
// CHECK1-NEXT:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP35]]
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[TMP36]] to [10 x [4 x %struct.S]]*
// CHECK1-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP38]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP39]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP40]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP41]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP42]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP43]], [[TMP44]]
// CHECK1-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP45]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP46:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX23:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i64 [[TMP46]]
// CHECK1-NEXT:    [[TMP47:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP47]] to i64
// CHECK1-NEXT:    [[ARRAYIDX24:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX23]], i64 [[IDXPROM]]
// CHECK1-NEXT:    [[TMP48:%.*]] = load i32, i32* [[ARRAYIDX24]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP48]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[ARRAYIDX24]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP49:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP49]], 1
// CHECK1-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP50:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP51:%.*]] = load i32, i32* [[TMP50]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP51]])
// CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP53:%.*]] = bitcast i32* [[VLA7]] to i8*
// CHECK1-NEXT:    store i8* [[TMP53]], i8** [[TMP52]], align 8
// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP55:%.*]] = inttoptr i64 [[TMP13]] to i8*
// CHECK1-NEXT:    store i8* [[TMP55]], i8** [[TMP54]], align 8
// CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP57:%.*]] = bitcast %struct.S* [[VLA15]] to i8*
// CHECK1-NEXT:    store i8* [[TMP57]], i8** [[TMP56]], align 8
// CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP59:%.*]] = inttoptr i64 [[TMP28]] to i8*
// CHECK1-NEXT:    store i8* [[TMP59]], i8** [[TMP58]], align 8
// CHECK1-NEXT:    [[TMP60:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP61:%.*]] = load i32, i32* [[TMP60]], align 4
// CHECK1-NEXT:    [[TMP62:%.*]] = bitcast [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP63:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP61]], i32 2, i64 32, i8* [[TMP62]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP63]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 [[TMP13]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP64]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE30:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[VLA7]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP65:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], align 4
// CHECK1-NEXT:    [[TMP66:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP65]], [[TMP66]]
// CHECK1-NEXT:    store i32 [[ADD27]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE29:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP64]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE29]], label [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done30:
// CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX9]], i64 [[TMP28]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY31:%.*]] = icmp eq %struct.S* [[ARRAYIDX9]], [[TMP67]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY31]], label [[OMP_ARRAYCPY_DONE38:%.*]], label [[OMP_ARRAYCPY_BODY32:%.*]]
// CHECK1:       omp.arraycpy.body32:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST33:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYCPY_DONE30]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT36:%.*]], [[OMP_ARRAYCPY_BODY32]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST34:%.*]] = phi %struct.S* [ [[ARRAYIDX9]], [[OMP_ARRAYCPY_DONE30]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT35:%.*]], [[OMP_ARRAYCPY_BODY32]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST34]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST33]])
// CHECK1-NEXT:    [[TMP68:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST34]] to i8*
// CHECK1-NEXT:    [[TMP69:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP68]], i8* align 4 [[TMP69]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT35]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST34]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT36]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST33]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE37:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT35]], [[TMP67]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE37]], label [[OMP_ARRAYCPY_DONE38]], label [[OMP_ARRAYCPY_BODY32]]
// CHECK1:       omp.arraycpy.done38:
// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP61]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 [[TMP13]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY39:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP70]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY39]], label [[OMP_ARRAYCPY_DONE46:%.*]], label [[OMP_ARRAYCPY_BODY40:%.*]]
// CHECK1:       omp.arraycpy.body40:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST41:%.*]] = phi i32* [ [[VLA7]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT44:%.*]], [[OMP_ARRAYCPY_BODY40]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST42:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT43:%.*]], [[OMP_ARRAYCPY_BODY40]] ]
// CHECK1-NEXT:    [[TMP71:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST41]], align 4
// CHECK1-NEXT:    [[TMP72:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST42]], i32 [[TMP71]] monotonic, align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT43]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST42]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT44]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST41]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE45:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT43]], [[TMP70]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE45]], label [[OMP_ARRAYCPY_DONE46]], label [[OMP_ARRAYCPY_BODY40]]
// CHECK1:       omp.arraycpy.done46:
// CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX9]], i64 [[TMP28]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY47:%.*]] = icmp eq %struct.S* [[ARRAYIDX9]], [[TMP73]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY47]], label [[OMP_ARRAYCPY_DONE55:%.*]], label [[OMP_ARRAYCPY_BODY48:%.*]]
// CHECK1:       omp.arraycpy.body48:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST49:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYCPY_DONE46]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT53:%.*]], [[OMP_ARRAYCPY_BODY48]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST50:%.*]] = phi %struct.S* [ [[ARRAYIDX9]], [[OMP_ARRAYCPY_DONE46]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT52:%.*]], [[OMP_ARRAYCPY_BODY48]] ]
// CHECK1-NEXT:    [[TMP74:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP75:%.*]] = load i32, i32* [[TMP74]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP75]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL51:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST50]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST49]])
// CHECK1-NEXT:    [[TMP76:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST50]] to i8*
// CHECK1-NEXT:    [[TMP77:%.*]] = bitcast %struct.S* [[CALL51]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP76]], i8* align 4 [[TMP77]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP75]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT52]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST50]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT53]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST49]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE54:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT52]], [[TMP73]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE54]], label [[OMP_ARRAYCPY_DONE55]], label [[OMP_ARRAYCPY_BODY48]]
// CHECK1:       omp.arraycpy.done55:
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP28]]
// CHECK1-NEXT:    [[ARRAYDESTROY_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA15]], [[TMP78]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_ISEMPTY]], label [[ARRAYDESTROY_DONE56:%.*]], label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP78]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[VLA15]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE56]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done56:
// CHECK1-NEXT:    [[TMP79:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP79]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [4 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [4 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S*
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S*
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = ptrtoint i8* [[TMP22]] to i64
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr i32, i32* [[TMP11]], i64 [[TMP14]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP24]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP24]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP20]], i64 [[TMP23]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY3:%.*]] = icmp eq %struct.S* [[TMP20]], [[TMP27]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY3]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY4:%.*]]
// CHECK1:       omp.arraycpy.body4:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST5:%.*]] = phi %struct.S* [ [[TMP17]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi %struct.S* [ [[TMP20]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT7:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST6]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST5]])
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]] to i8*
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP28]], i8* align 4 [[TMP29]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT7]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST5]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT7]], [[TMP27]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY4]]
// CHECK1:       omp.arraycpy.done10:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARR:%.*]], [10 x [4 x %struct.S]]* nonnull align 4 dereferenceable(160) [[ARRS:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[ARR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[ARRS_ADDR:%.*]] = alloca [10 x [4 x %struct.S]]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[ARRS4:%.*]] = alloca [10 x [4 x %struct.S]], align 16
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i32* [[ARR]], i32** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    store [10 x [4 x %struct.S]]* [[ARRS]], [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load [10 x [4 x %struct.S]]*, [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP0]], [[TMP1]]
// CHECK1-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
// CHECK1-NEXT:    [[TMP6:%.*]] = udiv exact i64 [[TMP5]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[VLA3:%.*]] = alloca i32, i64 [[TMP6]], align 16
// CHECK1-NEXT:    store i64 [[TMP6]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[VLA3]], i64 [[TMP6]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[VLA3]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[VLA3]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS4]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY5:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY5]], label [[OMP_ARRAYINIT_DONE10:%.*]], label [[OMP_ARRAYINIT_BODY6:%.*]]
// CHECK1:       omp.arrayinit.body6:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST7:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[OMP_ARRAYINIT_DONE]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYINIT_BODY6]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST7]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP9]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYINIT_DONE10]], label [[OMP_ARRAYINIT_BODY6]]
// CHECK1:       omp.arrayinit.done10:
// CHECK1-NEXT:    [[LHS_BEGIN:%.*]] = bitcast [10 x [4 x %struct.S]]* [[TMP3]] to %struct.S*
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [10 x [4 x %struct.S]]* [[ARRS4]] to %struct.S*
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA3]], i64 [[TMP18]]
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 [[IDXPROM]]
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[ARRAYIDX12]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK1-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i32* [[VLA3]] to i8*
// CHECK1-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP27:%.*]] = inttoptr i64 [[TMP6]] to i8*
// CHECK1-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP31]], i32 2, i64 24, i8* [[TMP32]], void (i8*, i8*)* @.omp.reduction.reduction_func.4, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP33]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr i32, i32* [[TMP2]], i64 [[TMP6]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP2]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[VLA3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi i32* [ [[TMP2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP35]], [[TMP36]]
// CHECK1-NEXT:    store i32 [[ADD15]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done18:
// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP37]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE26:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
// CHECK1:       omp.arraycpy.body20:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[OMP_ARRAYCPY_DONE18]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[OMP_ARRAYCPY_DONE18]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST22]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST21]])
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] to i8*
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP38]], i8* align 4 [[TMP39]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT24]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE25:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP37]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_BODY20]]
// CHECK1:       omp.arraycpy.done26:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP31]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr i32, i32* [[TMP2]], i64 [[TMP6]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY27:%.*]] = icmp eq i32* [[TMP2]], [[TMP40]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY27]], label [[OMP_ARRAYCPY_DONE34:%.*]], label [[OMP_ARRAYCPY_BODY28:%.*]]
// CHECK1:       omp.arraycpy.body28:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST29:%.*]] = phi i32* [ [[VLA3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT32:%.*]], [[OMP_ARRAYCPY_BODY28]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST30:%.*]] = phi i32* [ [[TMP2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT31:%.*]], [[OMP_ARRAYCPY_BODY28]] ]
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST29]], align 4
// CHECK1-NEXT:    [[TMP42:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST30]], i32 [[TMP41]] monotonic, align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT31]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST30]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT32]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST29]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE33:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT31]], [[TMP40]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE33]], label [[OMP_ARRAYCPY_DONE34]], label [[OMP_ARRAYCPY_BODY28]]
// CHECK1:       omp.arraycpy.done34:
// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY35:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP43]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY35]], label [[OMP_ARRAYCPY_DONE43:%.*]], label [[OMP_ARRAYCPY_BODY36:%.*]]
// CHECK1:       omp.arraycpy.body36:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST37:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[OMP_ARRAYCPY_DONE34]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT41:%.*]], [[OMP_ARRAYCPY_BODY36]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST38:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[OMP_ARRAYCPY_DONE34]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT40:%.*]], [[OMP_ARRAYCPY_BODY36]] ]
// CHECK1-NEXT:    [[TMP44:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[TMP44]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP45]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL39:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST38]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST37]])
// CHECK1-NEXT:    [[TMP46:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST38]] to i8*
// CHECK1-NEXT:    [[TMP47:%.*]] = bitcast %struct.S* [[CALL39]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP46]], i8* align 4 [[TMP47]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP45]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT40]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST38]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT41]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST37]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE42:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT40]], [[TMP43]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE42]], label [[OMP_ARRAYCPY_DONE43]], label [[OMP_ARRAYCPY_BODY36]]
// CHECK1:       omp.arraycpy.done43:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP31]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN44:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS4]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN44]], i64 40
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP48]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN44]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE45:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done45:
// CHECK1-NEXT:    [[TMP49:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP49]])
// CHECK1-NEXT:    [[TMP50:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP51:%.*]] = load i32, i32* [[TMP50]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP51]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.4
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S*
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S*
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP11]], i64 [[TMP14]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP21]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP20]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY3:%.*]] = icmp eq %struct.S* [[TMP20]], [[TMP24]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY3]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY4:%.*]]
// CHECK1:       omp.arraycpy.body4:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST5:%.*]] = phi %struct.S* [ [[TMP17]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi %struct.S* [ [[TMP20]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT7:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST6]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST5]])
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]] to i8*
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP25]], i8* align 4 [[TMP26]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT7]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST5]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT7]], [[TMP24]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY4]]
// CHECK1:       omp.arraycpy.done10:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARR:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[ARR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[ARR6:%.*]] = alloca [1 x [2 x i32]], align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    store i32* [[ARR]], i32** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP3:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP3]]
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 0
// CHECK1-NEXT:    [[TMP4:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP4]]
// CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX4]], i64 1
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [1 x [2 x i32]], [1 x [2 x i32]]* [[ARR6]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP6:%.*]] = ptrtoint i32* [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint i32* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    [[TMP9:%.*]] = sdiv exact i64 [[TMP8]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP10:%.*]] = bitcast [1 x [2 x i32]]* [[ARR6]] to i32*
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP10]], i64 [[TMP9]]
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [2 x i32]]* [[ARR6]] to i32*
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP14]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i64 [[TMP20]]
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX8]], i64 [[IDXPROM]]
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[ARRAYIDX9]], align 4
// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[INC]], i32* [[ARRAYIDX9]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
// CHECK1-NEXT:    [[TMP30:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP29]], i32 1, i64 8, i8* [[TMP30]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP31]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP32]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE15:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT13:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 4
// CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT13]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE14:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT13]], [[TMP32]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_DONE15]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done15:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY16:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP35]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY16]], label [[OMP_ARRAYCPY_DONE23:%.*]], label [[OMP_ARRAYCPY_BODY17:%.*]]
// CHECK1:       omp.arraycpy.body17:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST18:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT21:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST19:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST18]], align 4
// CHECK1-NEXT:    [[TMP37:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST19]], i32 [[TMP36]] monotonic, align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST19]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT21]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST18]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE22:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP35]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_DONE23]], label [[OMP_ARRAYCPY_BODY17]]
// CHECK1:       omp.arraycpy.done23:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP38]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP39]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[_TMP4:%.*]] = alloca %struct.S**, align 8
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 0
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
// CHECK1-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP6:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK1-NEXT:    [[TMP7:%.*]] = sub i64 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    [[TMP8:%.*]] = sdiv exact i64 [[TMP7]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP9:%.*]] = add nuw i64 [[TMP8]], 1
// CHECK1-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP11:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP11]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP9]], align 16
// CHECK1-NEXT:    store i64 [[TMP9]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP9]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP13:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = load %struct.S*, %struct.S** [[TMP13]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[TMP14]] to i64
// CHECK1-NEXT:    [[TMP16:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK1-NEXT:    [[TMP17:%.*]] = sub i64 [[TMP15]], [[TMP16]]
// CHECK1-NEXT:    [[TMP18:%.*]] = sdiv exact i64 [[TMP17]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP18]]
// CHECK1-NEXT:    store %struct.S** [[_TMP5]], %struct.S*** [[_TMP4]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP19]], %struct.S** [[_TMP5]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP23]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[VLA]] to i8*
// CHECK1-NEXT:    store i8* [[TMP32]], i8** [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP9]] to i8*
// CHECK1-NEXT:    store i8* [[TMP34]], i8** [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], i32 1, i64 16, i8* [[TMP37]], void (i8*, i8*)* @.omp.reduction.reduction_func.8, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP38]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 [[TMP9]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP39]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE11:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST8:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT9:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST8]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP40:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]] to i8*
// CHECK1-NEXT:    [[TMP41:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT9]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE10:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT9]], [[TMP39]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done11:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 [[TMP9]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY12:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP42]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY12]], label [[OMP_ARRAYCPY_DONE20:%.*]], label [[OMP_ARRAYCPY_BODY13:%.*]]
// CHECK1:       omp.arraycpy.body13:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST14:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK1-NEXT:    [[TMP43:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[TMP43]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL16:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST15]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST14]])
// CHECK1-NEXT:    [[TMP45:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]] to i8*
// CHECK1-NEXT:    [[TMP46:%.*]] = bitcast %struct.S* [[CALL16]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP45]], i8* align 4 [[TMP46]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST14]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE19:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP42]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_BODY13]]
// CHECK1:       omp.arraycpy.done20:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP9]]
// CHECK1-NEXT:    [[ARRAYDESTROY_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP47]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_ISEMPTY]], label [[ARRAYDESTROY_DONE21:%.*]], label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP47]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[VLA]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE21]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done21:
// CHECK1-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK1-NEXT:    [[TMP49:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = load i32, i32* [[TMP49]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP50]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.8
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 [[TMP14]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP16]], i8* align 4 [[TMP17]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR24:%.*]] = alloca [1 x [6 x %struct.S]], align 16
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca %struct.S**, align 8
// CHECK1-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 1
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 1
// CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
// CHECK1-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.10, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE12:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST9:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT10:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST9]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT10]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE11:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT10]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_DONE12]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done12:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY13:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY13]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY14:%.*]]
// CHECK1:       omp.arraycpy.body14:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST15:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL17:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST16]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST15]])
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]] to i8*
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL17]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST15]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT18]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY14]]
// CHECK1:       omp.arraycpy.done21:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN22:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN22]], i64 6
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN22]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE23:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done23:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.10
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR24:%.*]] = alloca [1 x [6 x %struct.S]], align 16
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca %struct.S**, align 8
// CHECK1-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 1
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 1
// CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
// CHECK1-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.12, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE12:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST9:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT10:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST9]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT10]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE11:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT10]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_DONE12]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done12:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY13:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY13]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY14:%.*]]
// CHECK1:       omp.arraycpy.body14:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST15:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL17:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST16]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST15]])
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]] to i8*
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL17]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST15]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT18]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY14]]
// CHECK1:       omp.arraycpy.done21:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN22:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN22]], i64 6
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN22]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE23:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done23:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.12
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 6
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR24:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca %struct.S**, align 8
// CHECK1-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 1
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP2]], i64 1
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 1
// CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
// CHECK1-NEXT:    [[TMP5:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S*, %struct.S** [[TMP5]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
// CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VAR24]], i64 [[TMP10]]
// CHECK1-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
// CHECK1-NEXT:    store %struct.S* [[TMP11]], %struct.S** [[_TMP6]], align 8
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP14]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]])
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast %struct.S* [[VAR24]] to i8*
// CHECK1-NEXT:    store i8* [[TMP22]], i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], i32 1, i64 8, i8* [[TMP23]], void (i8*, i8*)* @.omp.reduction.reduction_func.14, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYIDX1]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast %struct.S* [[ARRAYIDX1]] to i8*
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP25]], i8* align 4 [[TMP26]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYIDX1]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[ARRAYIDX1]] to i8*
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[CALL9]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP27]], i8* align 4 [[TMP28]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR24]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP13]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.14
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP11]], %struct.S* nonnull align 4 dereferenceable(4) [[TMP8]])
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP11]] to i8*
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [5 x %struct.S]* nonnull align 4 dereferenceable(20) [[VVAR2:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VVAR2_ADDR:%.*]] = alloca [5 x %struct.S]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VVAR22:%.*]] = alloca [5 x %struct.S], align 16
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [5 x %struct.S]* [[VVAR2]], [5 x %struct.S]** [[VVAR2_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [5 x %struct.S]*, [5 x %struct.S]** [[VVAR2_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[TMP0]], i64 0, i64 0
// CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[TMP0]], i64 0, i64 4
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR22]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 5
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP1]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP2:%.*]] = bitcast [5 x %struct.S]* [[TMP0]] to %struct.S*
// CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint %struct.S* [[TMP2]] to i64
// CHECK1-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
// CHECK1-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP7:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP7]], i64 [[TMP6]]
// CHECK1-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[TMP8]] to [5 x %struct.S]*
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK1-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP22]], i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    [[TMP25:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP24]], i32 1, i64 8, i8* [[TMP25]], void (i8*, i8*)* @.omp.reduction.reduction_func.16, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP26]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 5
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP27]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE8:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST5:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT6:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST5]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST5]] to i8*
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP28]], i8* align 4 [[TMP29]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT6]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST5]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE7:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT6]], [[TMP27]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE7]], label [[OMP_ARRAYCPY_DONE8]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done8:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP24]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 5
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY9:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP30]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY9]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY10:%.*]]
// CHECK1:       omp.arraycpy.body10:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST11:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY10]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT14:%.*]], [[OMP_ARRAYCPY_BODY10]] ]
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP32]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL13:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST12]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST11]])
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST12]] to i8*
// CHECK1-NEXT:    [[TMP34:%.*]] = bitcast %struct.S* [[CALL13]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP33]], i8* align 4 [[TMP34]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP32]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT14]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT15]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST11]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT14]], [[TMP30]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY10]]
// CHECK1:       omp.arraycpy.done17:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP24]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN18:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR22]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN18]], i64 5
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP35]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN18]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE19:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done19:
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP37]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.16
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 5
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR34:%.*]] = alloca [2 x %struct.S], align 4
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 2
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP4]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
// CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
// CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.18, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE11:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST8:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT9:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST8]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT9]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE10:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT9]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done11:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY12:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY12]], label [[OMP_ARRAYCPY_DONE20:%.*]], label [[OMP_ARRAYCPY_BODY13:%.*]]
// CHECK1:       omp.arraycpy.body13:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST14:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL16:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST15]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST14]])
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]] to i8*
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL16]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST14]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE19:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_BODY13]]
// CHECK1:       omp.arraycpy.done20:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN21:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN21]], i64 2
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN21]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE22:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done22:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.18
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR34:%.*]] = alloca [2 x %struct.S], align 4
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP4]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
// CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
// CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.20, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE11:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST8:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT9:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST8]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT9]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE10:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT9]], [[TMP31]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done11:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY12:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY12]], label [[OMP_ARRAYCPY_DONE20:%.*]], label [[OMP_ARRAYCPY_BODY13:%.*]]
// CHECK1:       omp.arraycpy.body13:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST14:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL16:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST15]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST14]])
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]] to i8*
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL16]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST14]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE19:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP34]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_BODY13]]
// CHECK1:       omp.arraycpy.done20:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN21:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN21]], i64 2
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN21]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE22:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done22:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.20
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 2
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[_TMP4:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
// CHECK1-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP6:%.*]] = sub i64 [[TMP4]], [[TMP5]]
// CHECK1-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S:%.*]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP8:%.*]] = add nuw i64 [[TMP7]], 1
// CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP10:%.*]] = call i8* @llvm.stacksave()
// CHECK1-NEXT:    store i8* [[TMP10]], i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP8]], align 16
// CHECK1-NEXT:    store i64 [[TMP8]], i64* [[__VLA_EXPR0]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP8]]
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP11]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP11]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP12:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast [4 x %struct.S]* [[TMP12]] to %struct.S*
// CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint %struct.S* [[TMP13]] to i64
// CHECK1-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP17]]
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast %struct.S* [[TMP18]] to [4 x %struct.S]*
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP19]], [4 x %struct.S]** [[_TMP4]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP23]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[VLA]] to i8*
// CHECK1-NEXT:    store i8* [[TMP32]], i8** [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP8]] to i8*
// CHECK1-NEXT:    store i8* [[TMP34]], i8** [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK1-NEXT:    [[TMP37:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], i32 1, i64 16, i8* [[TMP37]], void (i8*, i8*)* @.omp.reduction.reduction_func.22, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP38]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 [[TMP8]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP39]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST7:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST7]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP40:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]] to i8*
// CHECK1-NEXT:    [[TMP41:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP39]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done10:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 [[TMP8]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP42]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
// CHECK1:       omp.arraycpy.body12:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK1-NEXT:    [[TMP43:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[TMP43]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL15:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST14]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST13]])
// CHECK1-NEXT:    [[TMP45:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]] to i8*
// CHECK1-NEXT:    [[TMP46:%.*]] = bitcast %struct.S* [[CALL15]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP45]], i8* align 4 [[TMP46]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP42]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY12]]
// CHECK1:       omp.arraycpy.done19:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP8]]
// CHECK1-NEXT:    [[ARRAYDESTROY_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP47]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_ISEMPTY]], label [[ARRAYDESTROY_DONE20:%.*]], label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP47]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[VLA]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE20]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done20:
// CHECK1-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK1-NEXT:    [[TMP49:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP50:%.*]] = load i32, i32* [[TMP49]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP50]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.22
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 [[TMP14]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP16]], i8* align 4 [[TMP17]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR33:%.*]] = alloca [4 x %struct.S], align 16
// CHECK1-NEXT:    [[_TMP4:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK1-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[VAR33]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 4
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP3]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    store [4 x %struct.S]* [[VAR33]], [4 x %struct.S]** [[_TMP4]], align 8
// CHECK1-NEXT:    [[LHS_BEGIN:%.*]] = bitcast [4 x %struct.S]* [[TMP2]] to %struct.S*
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [4 x %struct.S]* [[VAR33]] to %struct.S*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP14]])
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP16:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 1, i64 8, i8* [[TMP19]], void (i8*, i8*)* @.omp.reduction.reduction_func.24, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP20]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP21]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST7:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST7]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]] to i8*
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP22]], i8* align 4 [[TMP23]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP21]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done10:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP24]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
// CHECK1:       omp.arraycpy.body12:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP26]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL15:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST14]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST13]])
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]] to i8*
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[CALL15]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP27]], i8* align 4 [[TMP28]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP26]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP24]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY12]]
// CHECK1:       omp.arraycpy.done19:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN20:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[VAR33]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN20]], i64 4
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP29]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN20]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE21:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done21:
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP31]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.24
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 4
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done2:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiLi42EET_v
// CHECK1-SAME: () #[[ATTR7]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[T:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK1-NEXT:    [[VAR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[VAR1:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK1-NEXT:    [[ARR:%.*]] = alloca [42 x %struct.S.0], align 16
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK1-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiLi42EET_v.vec to i8*), i64 8, i1 false)
// CHECK1-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i64 0, i64 0
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 1)
// CHECK1-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYINIT_BEGIN]], i64 1
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 2)
// CHECK1-NEXT:    store %struct.S.0* [[TEST]], %struct.S.0** [[VAR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR1]])
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[ARR]], i32 0, i32 0
// CHECK1-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 42
// CHECK1-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK1:       arrayctor.loop:
// CHECK1-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK1-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYCTOR_CUR]], i64 1
// CHECK1-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK1-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK1:       arrayctor.cont:
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.S.0*, %struct.S.0*, i32*, [2 x i32]*, [2 x %struct.S.0]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), i32* [[T_VAR]], %struct.S.0* [[TMP1]], %struct.S.0* [[VAR1]], i32* [[T_VAR1]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]])
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i32* [[T_VAR]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP2]])
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [42 x %struct.S.0]*, [2 x i32]*, i32*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [42 x %struct.S.0]* [[ARR]], [2 x i32]* [[VEC]], i32* [[T_VAR]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP3]])
// CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK1-NEXT:    [[ARRAY_BEGIN1:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[ARR]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN1]], i64 42
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP4]], [[ARRAYCTOR_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN1]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done2:
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR1]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAY_BEGIN3:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN3]], i64 2
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY4:%.*]]
// CHECK1:       arraydestroy.body4:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST5:%.*]] = phi %struct.S.0* [ [[TMP5]], [[ARRAYDESTROY_DONE2]] ], [ [[ARRAYDESTROY_ELEMENT6:%.*]], [[ARRAYDESTROY_BODY4]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT6]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST5]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT6]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE7:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT6]], [[ARRAY_BEGIN3]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE7]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY4]]
// CHECK1:       arraydestroy.done8:
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR5]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK1-NEXT:    ret i32 [[TMP6]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = load volatile double, double* @g, align 8
// CHECK1-NEXT:    [[CONV:%.*]] = fptrunc double [[TMP0]] to float
// CHECK1-NEXT:    store float [[CONV]], float* [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = fpext float [[TMP0]] to double
// CHECK1-NEXT:    [[TMP1:%.*]] = load volatile double, double* @g, align 8
// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], [[TMP1]]
// CHECK1-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD]] to float
// CHECK1-NEXT:    store float [[CONV2]], float* [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK1-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK1-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK1-NEXT:    call void @_ZN1SIiEC2Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]], i32 [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR1:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR1:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[VAR1_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[T_VAR1_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK1-NEXT:    [[_TMP5:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[VAR16:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK1-NEXT:    [[T_VAR17:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [4 x i8*], align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK1-NEXT:    [[REF_TMP22:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[VAR1]], %struct.S.0** [[VAR1_ADDR]], align 8
// CHECK1-NEXT:    store i32* [[T_VAR1]], i32** [[T_VAR1_ADDR]], align 8
// CHECK1-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR1_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[T_VAR1_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP1]], %struct.S.0** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP6]], %struct.S.0** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK1-NEXT:    store %struct.S.0* [[VAR4]], %struct.S.0** [[_TMP5]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK1-NEXT:    store i32 2147483647, i32* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP10]], 1
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP4]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 [[TMP16]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP18:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP5]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP19]] to i64
// CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[TMP5]], i64 0, i64 [[IDXPROM9]]
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast %struct.S.0* [[ARRAYIDX10]] to i8*
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast %struct.S.0* [[TMP18]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP20]], i8* align 4 [[TMP21]], i64 4, i1 false)
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i32* [[T_VAR3]] to i8*
// CHECK1-NEXT:    store i8* [[TMP24]], i8** [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast %struct.S.0* [[VAR4]] to i8*
// CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S.0* [[VAR16]] to i8*
// CHECK1-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP30:%.*]] = bitcast i32* [[T_VAR17]] to i8*
// CHECK1-NEXT:    store i8* [[TMP30]], i8** [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = bitcast [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP32:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], i32 4, i64 32, i8* [[TMP31]], void (i8*, i8*)* @.omp.reduction.reduction_func.26, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP32]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
// CHECK1-NEXT:    store i32 [[ADD12]], i32* [[TMP0]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S.0* @_ZN1SIiEanERKS0_(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK1-NEXT:    [[TMP35:%.*]] = bitcast %struct.S.0* [[TMP7]] to i8*
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast %struct.S.0* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK1-NEXT:    [[CALL13:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[CALL13]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK1:       land.rhs:
// CHECK1-NEXT:    [[CALL14:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK1-NEXT:    [[TOBOOL15:%.*]] = icmp ne i32 [[CALL14]], 0
// CHECK1-NEXT:    br label [[LAND_END]]
// CHECK1:       land.end:
// CHECK1-NEXT:    [[TMP37:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL15]], [[LAND_RHS]] ]
// CHECK1-NEXT:    [[CONV:%.*]] = zext i1 [[TMP37]] to i32
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[CONV]])
// CHECK1-NEXT:    [[TMP38:%.*]] = bitcast %struct.S.0* [[TMP2]] to i8*
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP38]], i8* align 4 [[TMP39]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[CMP16:%.*]] = icmp slt i32 [[TMP40]], [[TMP41]]
// CHECK1-NEXT:    br i1 [[CMP16]], label [[COND_TRUE17:%.*]], label [[COND_FALSE18:%.*]]
// CHECK1:       cond.true17:
// CHECK1-NEXT:    [[TMP42:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK1-NEXT:    br label [[COND_END19:%.*]]
// CHECK1:       cond.false18:
// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[T_VAR17]], align 4
// CHECK1-NEXT:    br label [[COND_END19]]
// CHECK1:       cond.end19:
// CHECK1-NEXT:    [[COND20:%.*]] = phi i32 [ [[TMP42]], [[COND_TRUE17]] ], [ [[TMP43]], [[COND_FALSE18]] ]
// CHECK1-NEXT:    store i32 [[COND20]], i32* [[TMP3]], align 4
// CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP45:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP44]] monotonic, align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL21:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S.0* @_ZN1SIiEanERKS0_(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK1-NEXT:    [[TMP46:%.*]] = bitcast %struct.S.0* [[TMP7]] to i8*
// CHECK1-NEXT:    [[TMP47:%.*]] = bitcast %struct.S.0* [[CALL21]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP46]], i8* align 4 [[TMP47]], i64 4, i1 false)
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL23:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK1-NEXT:    [[TOBOOL24:%.*]] = icmp ne i32 [[CALL23]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL24]], label [[LAND_RHS25:%.*]], label [[LAND_END28:%.*]]
// CHECK1:       land.rhs25:
// CHECK1-NEXT:    [[CALL26:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK1-NEXT:    [[TOBOOL27:%.*]] = icmp ne i32 [[CALL26]], 0
// CHECK1-NEXT:    br label [[LAND_END28]]
// CHECK1:       land.end28:
// CHECK1-NEXT:    [[TMP48:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE2]] ], [ [[TOBOOL27]], [[LAND_RHS25]] ]
// CHECK1-NEXT:    [[CONV29:%.*]] = zext i1 [[TMP48]] to i32
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP22]], i32 [[CONV29]])
// CHECK1-NEXT:    [[TMP49:%.*]] = bitcast %struct.S.0* [[TMP2]] to i8*
// CHECK1-NEXT:    [[TMP50:%.*]] = bitcast %struct.S.0* [[REF_TMP22]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP49]], i8* align 4 [[TMP50]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP22]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[TMP51:%.*]] = load i32, i32* [[T_VAR17]], align 4
// CHECK1-NEXT:    [[TMP52:%.*]] = atomicrmw min i32* [[TMP3]], i32 [[TMP51]] monotonic, align 4
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR5]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.26
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [4 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [4 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK1-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP22]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8* [[TMP25]] to i32*
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 3
// CHECK1-NEXT:    [[TMP28:%.*]] = load i8*, i8** [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8* [[TMP28]] to i32*
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
// CHECK1-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S.0* @_ZN1SIiEanERKS0_(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP17]], %struct.S.0* nonnull align 4 dereferenceable(4) [[TMP14]])
// CHECK1-NEXT:    [[TMP32:%.*]] = bitcast %struct.S.0* [[TMP17]] to i8*
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S.0* [[CALL]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK1-NEXT:    [[CALL2:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP23]])
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[CALL2]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK1:       land.rhs:
// CHECK1-NEXT:    [[CALL3:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP20]])
// CHECK1-NEXT:    [[TOBOOL4:%.*]] = icmp ne i32 [[CALL3]], 0
// CHECK1-NEXT:    br label [[LAND_END]]
// CHECK1:       land.end:
// CHECK1-NEXT:    [[TMP34:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL4]], [[LAND_RHS]] ]
// CHECK1-NEXT:    [[CONV:%.*]] = zext i1 [[TMP34]] to i32
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[CONV]])
// CHECK1-NEXT:    [[TMP35:%.*]] = bitcast %struct.S.0* [[TMP23]] to i8*
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP26]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP37]], [[TMP38]]
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[TMP26]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP39]], [[COND_TRUE]] ], [ [[TMP40]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[TMP29]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEanERKS0_
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR7]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP0]], %struct.S.0** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret %struct.S.0* [[THIS1]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEcviEv
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) #[[ATTR7]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SIiED2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR5]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP3]], %struct.S.0** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP14]] to i64
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 [[TMP13]], i32* [[ARRAYIDX]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP16]] to i64
// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[TMP2]], i64 0, i64 [[IDXPROM5]]
// CHECK1-NEXT:    [[TMP17:%.*]] = bitcast %struct.S.0* [[ARRAYIDX6]] to i8*
// CHECK1-NEXT:    [[TMP18:%.*]] = bitcast %struct.S.0* [[TMP15]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP17]], i8* align 4 [[TMP18]], i64 4, i1 false)
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]])
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i32* [[T_VAR3]] to i8*
// CHECK1-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP23:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 1, i64 8, i8* [[TMP22]], void (i8*, i8*)* @.omp.reduction.reduction_func.28, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP23]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK1:       land.rhs:
// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TOBOOL8:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK1-NEXT:    br label [[LAND_END]]
// CHECK1:       land.end:
// CHECK1-NEXT:    [[TMP26:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL8]], [[LAND_RHS]] ]
// CHECK1-NEXT:    [[CONV:%.*]] = zext i1 [[TMP26]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TOBOOL9:%.*]] = icmp ne i32 [[TMP27]], 0
// CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, i32* [[TMP0]] monotonic, align 4
// CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
// CHECK1:       atomic_cont:
// CHECK1-NEXT:    [[TMP28:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP34:%.*]], [[LAND_END14:%.*]] ]
// CHECK1-NEXT:    store i32 [[TMP28]], i32* [[_TMP10]], align 4
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[_TMP10]], align 4
// CHECK1-NEXT:    [[TOBOOL11:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL11]], label [[LAND_RHS12:%.*]], label [[LAND_END14]]
// CHECK1:       land.rhs12:
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK1-NEXT:    [[TOBOOL13:%.*]] = icmp ne i32 [[TMP30]], 0
// CHECK1-NEXT:    br label [[LAND_END14]]
// CHECK1:       land.end14:
// CHECK1-NEXT:    [[TMP31:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL13]], [[LAND_RHS12]] ]
// CHECK1-NEXT:    [[CONV15:%.*]] = zext i1 [[TMP31]] to i32
// CHECK1-NEXT:    store i32 [[CONV15]], i32* [[ATOMIC_TEMP]], align 4
// CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[ATOMIC_TEMP]], align 4
// CHECK1-NEXT:    [[TMP33:%.*]] = cmpxchg i32* [[TMP0]], i32 [[TMP28]], i32 [[TMP32]] monotonic monotonic, align 4
// CHECK1-NEXT:    [[TMP34]] = extractvalue { i32, i1 } [[TMP33]], 0
// CHECK1-NEXT:    [[TMP35:%.*]] = extractvalue { i32, i1 } [[TMP33]], 1
// CHECK1-NEXT:    br i1 [[TMP35]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
// CHECK1:       atomic_exit:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP6]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.28
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK1:       land.rhs:
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK1-NEXT:    [[TOBOOL2:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br label [[LAND_END]]
// CHECK1:       land.end:
// CHECK1-NEXT:    [[TMP14:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
// CHECK1-NEXT:    [[CONV:%.*]] = zext i1 [[TMP14]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[TMP11]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [42 x %struct.S.0]* nonnull align 4 dereferenceable(168) [[ARR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[ARR_ADDR:%.*]] = alloca [42 x %struct.S.0]*, align 8
// CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[ARR4:%.*]] = alloca [40 x %struct.S.0], align 16
// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK1-NEXT:    [[REF_TMP20:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT:    store [42 x %struct.S.0]* [[ARR]], [42 x %struct.S.0]** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load [42 x %struct.S.0]*, [42 x %struct.S.0]** [[ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[TMP]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK1-NEXT:    store %struct.S.0* [[TMP5]], %struct.S.0** [[_TMP1]], align 8
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[TMP0]], i64 0, i64 1
// CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[TMP0]], i64 0, i64 40
// CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [40 x %struct.S.0], [40 x %struct.S.0]* [[ARR4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK1:       omp.arrayinit.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK1:       omp.arrayinit.done:
// CHECK1-NEXT:    [[TMP7:%.*]] = bitcast [42 x %struct.S.0]* [[TMP0]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S.0* [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S.0* [[ARRAYIDX]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S.0* getelementptr ([[STRUCT_S_0]], %struct.S.0* null, i32 1) to i64)
// CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[TMP12]], i64 [[TMP11]]
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP13]] to [42 x %struct.S.0]*
// CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP17]], 1
// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1:       cond.true:
// CHECK1-NEXT:    br label [[COND_END:%.*]]
// CHECK1:       cond.false:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK1:       omp.inner.for.cond.cleanup:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
// CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
// CHECK1-NEXT:    store i32 [[TMP23]], i32* [[ARRAYIDX6]], align 4
// CHECK1-NEXT:    [[TMP25:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
// CHECK1-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP26]] to i64
// CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[TMP3]], i64 0, i64 [[IDXPROM7]]
// CHECK1-NEXT:    [[TMP27:%.*]] = bitcast %struct.S.0* [[ARRAYIDX8]] to i8*
// CHECK1-NEXT:    [[TMP28:%.*]] = bitcast %struct.S.0* [[TMP25]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP27]], i8* align 4 [[TMP28]], i64 4, i1 false)
// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1:       omp.body.continue:
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP31]])
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP33:%.*]] = bitcast %struct.S.0* [[RHS_BEGIN]] to i8*
// CHECK1-NEXT:    store i8* [[TMP33]], i8** [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK1-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK1-NEXT:    [[TMP37:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP35]], i32 1, i64 8, i8* [[TMP36]], void (i8*, i8*)* @.omp.reduction.reduction_func.30, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    switch i32 [[TMP37]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK1-NEXT:    ]
// CHECK1:       .omp.reduction.case1:
// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAYIDX]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAYIDX]], [[TMP38]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE15:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi %struct.S.0* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT13:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST10]])
// CHECK1-NEXT:    [[CALL11:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[CALL]], [[CALL11]]
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[ADD12]])
// CHECK1-NEXT:    [[TMP39:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST10]] to i8*
// CHECK1-NEXT:    [[TMP40:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP39]], i8* align 4 [[TMP40]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT13]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE14:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT13]], [[TMP38]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_DONE15]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done15:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP35]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.case2:
// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAYIDX]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY16:%.*]] = icmp eq %struct.S.0* [[ARRAYIDX]], [[TMP41]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY16]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY17:%.*]]
// CHECK1:       omp.arraycpy.body17:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST18:%.*]] = phi %struct.S.0* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST19:%.*]] = phi %struct.S.0* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK1-NEXT:    [[TMP42:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[TMP42]], align 4
// CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP43]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[CALL21:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST19]])
// CHECK1-NEXT:    [[CALL22:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST18]])
// CHECK1-NEXT:    [[ADD23:%.*]] = add nsw i32 [[CALL21]], [[CALL22]]
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP20]], i32 [[ADD23]])
// CHECK1-NEXT:    [[TMP44:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST19]] to i8*
// CHECK1-NEXT:    [[TMP45:%.*]] = bitcast %struct.S.0* [[REF_TMP20]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP44]], i8* align 4 [[TMP45]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP20]]) #[[ATTR5]]
// CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP43]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST19]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST18]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP41]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY17]]
// CHECK1:       omp.arraycpy.done27:
// CHECK1-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP35]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK1:       .omp.reduction.default:
// CHECK1-NEXT:    [[ARRAY_BEGIN28:%.*]] = getelementptr inbounds [40 x %struct.S.0], [40 x %struct.S.0]* [[ARR4]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN28]], i64 40
// CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK1:       arraydestroy.body:
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP46]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN28]]
// CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE29:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK1:       arraydestroy.done29:
// CHECK1-NEXT:    [[TMP47:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP48:%.*]] = load i32, i32* [[TMP47]], align 4
// CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP48]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.30
// CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S.0*
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[TMP11]], i64 40
// CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK1:       omp.arraycpy.body:
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK1-NEXT:    [[CALL:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK1-NEXT:    [[CALL2:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CALL]], [[CALL2]]
// CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[ADD]])
// CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK1-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
// CHECK1:       omp.arraycpy.done3:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = load volatile double, double* @g, align 8
// CHECK1-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i32
// CHECK1-NEXT:    store i32 [[CONV]], i32* [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to double
// CHECK1-NEXT:    [[TMP1:%.*]] = load volatile double, double* @g, align 8
// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], [[TMP1]]
// CHECK1-NEXT:    [[CONV2:%.*]] = fptosi double [[ADD]] to i32
// CHECK1-NEXT:    store i32 [[CONV2]], i32* [[F]], align 4
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK1-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK1-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@main
// CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK2-NEXT:    [[T_VAR:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[T_VAR1:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK2-NEXT:    [[S_ARR:%.*]] = alloca [4 x %struct.S], align 16
// CHECK2-NEXT:    [[VAR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[VAR1:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK2-NEXT:    [[ARRS:%.*]] = alloca [10 x [4 x %struct.S]], align 16
// CHECK2-NEXT:    [[VAR2:%.*]] = alloca %struct.S**, align 8
// CHECK2-NEXT:    [[VVAR2:%.*]] = alloca [5 x %struct.S], align 16
// CHECK2-NEXT:    [[VAR3:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK2-NEXT:    store float 0.000000e+00, float* [[T_VAR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const.main.vec to i8*), i64 8, i1 false)
// CHECK2-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[S_ARR]], i64 0, i64 0
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], float 1.000000e+00)
// CHECK2-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_BEGIN]], i64 1
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float 2.000000e+00)
// CHECK2-NEXT:    [[ARRAYINIT_ELEMENT1:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_ELEMENT]], i64 1
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT1]], float 3.000000e+00)
// CHECK2-NEXT:    [[ARRAYINIT_ELEMENT2:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYINIT_ELEMENT1]], i64 1
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT2]], float 4.000000e+00)
// CHECK2-NEXT:    store %struct.S* [[TEST]], %struct.S** [[VAR]], align 8
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR1]])
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 40
// CHECK2-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK2:       arrayctor.loop:
// CHECK2-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK2-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYCTOR_CUR]], i64 1
// CHECK2-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq %struct.S* [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK2-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK2:       arrayctor.cont:
// CHECK2-NEXT:    [[CALL:%.*]] = call %struct.S** @_Z3foov()
// CHECK2-NEXT:    store %struct.S** [[CALL]], %struct.S*** [[VAR2]], align 8
// CHECK2-NEXT:    [[ARRAY_BEGIN3:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR2]], i32 0, i32 0
// CHECK2-NEXT:    [[ARRAYCTOR_END4:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN3]], i64 5
// CHECK2-NEXT:    br label [[ARRAYCTOR_LOOP5:%.*]]
// CHECK2:       arrayctor.loop5:
// CHECK2-NEXT:    [[ARRAYCTOR_CUR6:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN3]], [[ARRAYCTOR_CONT]] ], [ [[ARRAYCTOR_NEXT7:%.*]], [[ARRAYCTOR_LOOP5]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR6]])
// CHECK2-NEXT:    [[ARRAYCTOR_NEXT7]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYCTOR_CUR6]], i64 1
// CHECK2-NEXT:    [[ARRAYCTOR_DONE8:%.*]] = icmp eq %struct.S* [[ARRAYCTOR_NEXT7]], [[ARRAYCTOR_END4]]
// CHECK2-NEXT:    br i1 [[ARRAYCTOR_DONE8]], label [[ARRAYCTOR_CONT9:%.*]], label [[ARRAYCTOR_LOOP5]]
// CHECK2:       arrayctor.cont9:
// CHECK2-NEXT:    store [4 x %struct.S]* [[S_ARR]], [4 x %struct.S]** [[VAR3]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float*, %struct.S*, %struct.S*, float*, [2 x i32]*, [4 x %struct.S]*)* @.omp_outlined. to void (i32*, i32*, ...)*), float* [[T_VAR]], %struct.S* [[TMP1]], %struct.S* [[VAR1]], float* [[T_VAR1]], [2 x i32]* [[VEC]], [4 x %struct.S]* [[S_ARR]])
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[VEC]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
// CHECK2-NEXT:    [[TMP4:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP4]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = mul nuw i64 10, [[TMP3]]
// CHECK2-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 16
// CHECK2-NEXT:    store i64 [[TMP3]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, [2 x i32]*, [10 x [4 x %struct.S]]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 10, i64 [[TMP3]], i32* [[VLA]], [2 x i32]* [[VEC]], [10 x [4 x %struct.S]]* [[ARRS]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, [10 x [4 x %struct.S]]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 10, i64 [[TMP3]], i32* [[VLA]], [10 x [4 x %struct.S]]* [[ARRS]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 10, i64 [[TMP3]], i32* [[VLA]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S***)* @.omp_outlined..13 to void (i32*, i32*, ...)*), %struct.S*** [[VAR2]])
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [5 x %struct.S]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), [5 x %struct.S]* [[VVAR2]])
// CHECK2-NEXT:    [[TMP6:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP6]])
// CHECK2-NEXT:    [[TMP7:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..19 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP7]])
// CHECK2-NEXT:    [[TMP8:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP8]])
// CHECK2-NEXT:    [[TMP9:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [4 x %struct.S]*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), [4 x %struct.S]* [[TMP9]])
// CHECK2-NEXT:    [[CALL10:%.*]] = call i32 @_Z5tmainIiLi42EET_v()
// CHECK2-NEXT:    store i32 [[CALL10]], i32* [[RETVAL]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP10]])
// CHECK2-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR2]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN11]], i64 5
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ARRAYCTOR_CONT9]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5:[0-9]+]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done12:
// CHECK2-NEXT:    [[ARRAY_BEGIN13:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN13]], i64 40
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY14:%.*]]
// CHECK2:       arraydestroy.body14:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST15:%.*]] = phi %struct.S* [ [[TMP12]], [[ARRAYDESTROY_DONE12]] ], [ [[ARRAYDESTROY_ELEMENT16:%.*]], [[ARRAYDESTROY_BODY14]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT16]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST15]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT16]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE17:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT16]], [[ARRAY_BEGIN13]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE17]], label [[ARRAYDESTROY_DONE18:%.*]], label [[ARRAYDESTROY_BODY14]]
// CHECK2:       arraydestroy.done18:
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR1]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAY_BEGIN19:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[S_ARR]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN19]], i64 4
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY20:%.*]]
// CHECK2:       arraydestroy.body20:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST21:%.*]] = phi %struct.S* [ [[TMP13]], [[ARRAYDESTROY_DONE18]] ], [ [[ARRAYDESTROY_ELEMENT22:%.*]], [[ARRAYDESTROY_BODY20]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT22]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST21]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT22]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE23:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT22]], [[ARRAY_BEGIN19]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE23]], label [[ARRAYDESTROY_DONE24:%.*]], label [[ARRAYDESTROY_BODY20]]
// CHECK2:       arraydestroy.done24:
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR5]]
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK2-NEXT:    ret i32 [[TMP14]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void @_ZN1SIfEC2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK2-NEXT:    call void @_ZN1SIfEC2Ef(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]], float [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR1:%.*]], float* nonnull align 4 dereferenceable(4) [[T_VAR1:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[S_ARR:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[T_VAR_ADDR:%.*]] = alloca float*, align 8
// CHECK2-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[VAR1_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[T_VAR1_ADDR:%.*]] = alloca float*, align 8
// CHECK2-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK2-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[T_VAR3:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[VAR16:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK2-NEXT:    [[T_VAR17:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[_TMP22:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[REF_TMP25:%.*]] = alloca [[STRUCT_S]], align 4
// CHECK2-NEXT:    [[ATOMIC_TEMP35:%.*]] = alloca float, align 4
// CHECK2-NEXT:    [[_TMP36:%.*]] = alloca float, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store float* [[T_VAR]], float** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S* [[VAR]], %struct.S** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S* [[VAR1]], %struct.S** [[VAR1_ADDR]], align 8
// CHECK2-NEXT:    store float* [[T_VAR1]], float** [[T_VAR1_ADDR]], align 8
// CHECK2-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[S_ARR]], [4 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load float*, float** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S*, %struct.S** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[VAR1_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load float*, float** [[T_VAR1_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP1]], %struct.S** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S*, %struct.S** [[TMP]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP6]], %struct.S** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    store float 0.000000e+00, float* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[_TMP1]], align 8
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK2-NEXT:    store %struct.S* [[VAR4]], %struct.S** [[_TMP5]], align 8
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK2-NEXT:    store float 0x47EFFFFFE0000000, float* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP10]], 1
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = fptosi float [[TMP16]] to i32
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP4]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load %struct.S*, %struct.S** [[_TMP5]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP19]] to i64
// CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP5]], i64 0, i64 [[IDXPROM9]]
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast %struct.S* [[ARRAYIDX10]] to i8*
// CHECK2-NEXT:    [[TMP21:%.*]] = bitcast %struct.S* [[TMP18]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP20]], i8* align 4 [[TMP21]], i64 4, i1 false)
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP24:%.*]] = bitcast float* [[T_VAR3]] to i8*
// CHECK2-NEXT:    store i8* [[TMP24]], i8** [[TMP23]], align 8
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[VAR4]] to i8*
// CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[VAR16]] to i8*
// CHECK2-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP30:%.*]] = bitcast float* [[T_VAR17]] to i8*
// CHECK2-NEXT:    store i8* [[TMP30]], i8** [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP32:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 4, i64 32, i8* [[TMP31]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP32]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP33:%.*]] = load float, float* [[TMP0]], align 4
// CHECK2-NEXT:    [[TMP34:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = fadd float [[TMP33]], [[TMP34]]
// CHECK2-NEXT:    store float [[ADD12]], float* [[TMP0]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK2-NEXT:    [[CALL13:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[CALL13]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK2:       land.rhs:
// CHECK2-NEXT:    [[CALL14:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK2-NEXT:    [[TOBOOL15:%.*]] = fcmp une float [[CALL14]], 0.000000e+00
// CHECK2-NEXT:    br label [[LAND_END]]
// CHECK2:       land.end:
// CHECK2-NEXT:    [[TMP37:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL15]], [[LAND_RHS]] ]
// CHECK2-NEXT:    [[CONV16:%.*]] = uitofp i1 [[TMP37]] to float
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]], float [[CONV16]])
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[TMP2]] to i8*
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast %struct.S* [[REF_TMP]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP38]], i8* align 4 [[TMP39]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK2-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP3]], align 4
// CHECK2-NEXT:    [[TMP41:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[CMP17:%.*]] = fcmp olt float [[TMP40]], [[TMP41]]
// CHECK2-NEXT:    br i1 [[CMP17]], label [[COND_TRUE18:%.*]], label [[COND_FALSE19:%.*]]
// CHECK2:       cond.true18:
// CHECK2-NEXT:    [[TMP42:%.*]] = load float, float* [[TMP3]], align 4
// CHECK2-NEXT:    br label [[COND_END20:%.*]]
// CHECK2:       cond.false19:
// CHECK2-NEXT:    [[TMP43:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK2-NEXT:    br label [[COND_END20]]
// CHECK2:       cond.end20:
// CHECK2-NEXT:    [[COND21:%.*]] = phi float [ [[TMP42]], [[COND_TRUE18]] ], [ [[TMP43]], [[COND_FALSE19]] ]
// CHECK2-NEXT:    store float [[COND21]], float* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP44:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP45:%.*]] = bitcast float* [[TMP0]] to i32*
// CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, i32* [[TMP45]] monotonic, align 4
// CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
// CHECK2:       atomic_cont:
// CHECK2-NEXT:    [[TMP46:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP54:%.*]], [[ATOMIC_CONT]] ]
// CHECK2-NEXT:    [[TMP47:%.*]] = bitcast float* [[ATOMIC_TEMP]] to i32*
// CHECK2-NEXT:    [[TMP48:%.*]] = bitcast i32 [[TMP46]] to float
// CHECK2-NEXT:    store float [[TMP48]], float* [[_TMP22]], align 4
// CHECK2-NEXT:    [[TMP49:%.*]] = load float, float* [[_TMP22]], align 4
// CHECK2-NEXT:    [[TMP50:%.*]] = load float, float* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[ADD23:%.*]] = fadd float [[TMP49]], [[TMP50]]
// CHECK2-NEXT:    store float [[ADD23]], float* [[ATOMIC_TEMP]], align 4
// CHECK2-NEXT:    [[TMP51:%.*]] = load i32, i32* [[TMP47]], align 4
// CHECK2-NEXT:    [[TMP52:%.*]] = bitcast float* [[TMP0]] to i32*
// CHECK2-NEXT:    [[TMP53:%.*]] = cmpxchg i32* [[TMP52]], i32 [[TMP46]], i32 [[TMP51]] monotonic monotonic, align 4
// CHECK2-NEXT:    [[TMP54]] = extractvalue { i32, i1 } [[TMP53]], 0
// CHECK2-NEXT:    [[TMP55:%.*]] = extractvalue { i32, i1 } [[TMP53]], 1
// CHECK2-NEXT:    br i1 [[TMP55]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
// CHECK2:       atomic_exit:
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL24:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK2-NEXT:    [[TMP56:%.*]] = bitcast %struct.S* [[TMP7]] to i8*
// CHECK2-NEXT:    [[TMP57:%.*]] = bitcast %struct.S* [[CALL24]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP56]], i8* align 4 [[TMP57]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL26:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK2-NEXT:    [[TOBOOL27:%.*]] = fcmp une float [[CALL26]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[TOBOOL27]], label [[LAND_RHS28:%.*]], label [[LAND_END31:%.*]]
// CHECK2:       land.rhs28:
// CHECK2-NEXT:    [[CALL29:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK2-NEXT:    [[TOBOOL30:%.*]] = fcmp une float [[CALL29]], 0.000000e+00
// CHECK2-NEXT:    br label [[LAND_END31]]
// CHECK2:       land.end31:
// CHECK2-NEXT:    [[TMP58:%.*]] = phi i1 [ false, [[ATOMIC_EXIT]] ], [ [[TOBOOL30]], [[LAND_RHS28]] ]
// CHECK2-NEXT:    [[CONV32:%.*]] = uitofp i1 [[TMP58]] to float
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP25]], float [[CONV32]])
// CHECK2-NEXT:    [[TMP59:%.*]] = bitcast %struct.S* [[TMP2]] to i8*
// CHECK2-NEXT:    [[TMP60:%.*]] = bitcast %struct.S* [[REF_TMP25]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP59]], i8* align 4 [[TMP60]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP25]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[TMP61:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[TMP62:%.*]] = bitcast float* [[TMP3]] to i32*
// CHECK2-NEXT:    [[ATOMIC_LOAD33:%.*]] = load atomic i32, i32* [[TMP62]] monotonic, align 4
// CHECK2-NEXT:    br label [[ATOMIC_CONT34:%.*]]
// CHECK2:       atomic_cont34:
// CHECK2-NEXT:    [[TMP63:%.*]] = phi i32 [ [[ATOMIC_LOAD33]], [[LAND_END31]] ], [ [[TMP73:%.*]], [[COND_END40:%.*]] ]
// CHECK2-NEXT:    [[TMP64:%.*]] = bitcast float* [[ATOMIC_TEMP35]] to i32*
// CHECK2-NEXT:    [[TMP65:%.*]] = bitcast i32 [[TMP63]] to float
// CHECK2-NEXT:    store float [[TMP65]], float* [[_TMP36]], align 4
// CHECK2-NEXT:    [[TMP66:%.*]] = load float, float* [[_TMP36]], align 4
// CHECK2-NEXT:    [[TMP67:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[CMP37:%.*]] = fcmp olt float [[TMP66]], [[TMP67]]
// CHECK2-NEXT:    br i1 [[CMP37]], label [[COND_TRUE38:%.*]], label [[COND_FALSE39:%.*]]
// CHECK2:       cond.true38:
// CHECK2-NEXT:    [[TMP68:%.*]] = load float, float* [[_TMP36]], align 4
// CHECK2-NEXT:    br label [[COND_END40]]
// CHECK2:       cond.false39:
// CHECK2-NEXT:    [[TMP69:%.*]] = load float, float* [[T_VAR17]], align 4
// CHECK2-NEXT:    br label [[COND_END40]]
// CHECK2:       cond.end40:
// CHECK2-NEXT:    [[COND41:%.*]] = phi float [ [[TMP68]], [[COND_TRUE38]] ], [ [[TMP69]], [[COND_FALSE39]] ]
// CHECK2-NEXT:    store float [[COND41]], float* [[ATOMIC_TEMP35]], align 4
// CHECK2-NEXT:    [[TMP70:%.*]] = load i32, i32* [[TMP64]], align 4
// CHECK2-NEXT:    [[TMP71:%.*]] = bitcast float* [[TMP3]] to i32*
// CHECK2-NEXT:    [[TMP72:%.*]] = cmpxchg i32* [[TMP71]], i32 [[TMP63]], i32 [[TMP70]] monotonic monotonic, align 4
// CHECK2-NEXT:    [[TMP73]] = extractvalue { i32, i1 } [[TMP72]], 0
// CHECK2-NEXT:    [[TMP74:%.*]] = extractvalue { i32, i1 } [[TMP72]], 1
// CHECK2-NEXT:    br i1 [[TMP74]], label [[ATOMIC_EXIT42:%.*]], label [[ATOMIC_CONT34]]
// CHECK2:       atomic_exit42:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR16]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP9]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [4 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [4 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to float*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to float*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to %struct.S*
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S*
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S*
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP22]] to %struct.S*
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i8* [[TMP25]] to float*
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP28:%.*]] = load i8*, i8** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8* [[TMP28]] to float*
// CHECK2-NEXT:    [[TMP30:%.*]] = load float, float* [[TMP11]], align 4
// CHECK2-NEXT:    [[TMP31:%.*]] = load float, float* [[TMP8]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = fadd float [[TMP30]], [[TMP31]]
// CHECK2-NEXT:    store float [[ADD]], float* [[TMP11]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP17]], %struct.S* nonnull align 4 dereferenceable(4) [[TMP14]])
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[TMP17]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK2-NEXT:    [[CALL2:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP23]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[CALL2]], 0.000000e+00
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK2:       land.rhs:
// CHECK2-NEXT:    [[CALL3:%.*]] = call float @_ZN1SIfEcvfEv(%struct.S* nonnull align 4 dereferenceable(4) [[TMP20]])
// CHECK2-NEXT:    [[TOBOOL4:%.*]] = fcmp une float [[CALL3]], 0.000000e+00
// CHECK2-NEXT:    br label [[LAND_END]]
// CHECK2:       land.end:
// CHECK2-NEXT:    [[TMP34:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL4]], [[LAND_RHS]] ]
// CHECK2-NEXT:    [[CONV:%.*]] = uitofp i1 [[TMP34]] to float
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ef(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]], float [[CONV]])
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast %struct.S* [[TMP23]] to i8*
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast %struct.S* [[REF_TMP]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK2-NEXT:    [[TMP37:%.*]] = load float, float* [[TMP29]], align 4
// CHECK2-NEXT:    [[TMP38:%.*]] = load float, float* [[TMP26]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = fcmp olt float [[TMP37]], [[TMP38]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP39:%.*]] = load float, float* [[TMP29]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP26]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi float [ [[TMP39]], [[COND_TRUE]] ], [ [[TMP40]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store float [[COND]], float* [[TMP29]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfEanERKS0_
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], %struct.S* nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR7:[0-9]+]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP0]], %struct.S** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    ret %struct.S* [[THIS1]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfEcvfEv
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) #[[ATTR7]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    ret float 0.000000e+00
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void @_ZN1SIfED2Ev(%struct.S* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR5]]
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [10 x [4 x %struct.S]]* nonnull align 4 dereferenceable(160) [[ARRS:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[ARR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK2-NEXT:    [[ARRS_ADDR:%.*]] = alloca [10 x [4 x %struct.S]]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i32* [[ARR]], i32** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    store [10 x [4 x %struct.S]]* [[ARRS]], [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load [10 x [4 x %struct.S]]*, [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP5]]
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 0
// CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
// CHECK2-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
// CHECK2-NEXT:    [[TMP8:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP8]]
// CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX5]], i64 [[LB_ADD_LEN]]
// CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint i32* [[ARRAYIDX6]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = ptrtoint i32* [[ARRAYIDX3]] to i64
// CHECK2-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    [[TMP12:%.*]] = sdiv exact i64 [[TMP11]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP13:%.*]] = add nuw i64 [[TMP12]], 1
// CHECK2-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP15:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP15]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[VLA7:%.*]] = alloca i32, i64 [[TMP13]], align 16
// CHECK2-NEXT:    store i64 [[TMP13]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr i32, i32* [[VLA7]], i64 [[TMP13]]
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[VLA7]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[VLA7]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP17:%.*]] = ptrtoint i32* [[TMP2]] to i64
// CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i32* [[ARRAYIDX3]] to i64
// CHECK2-NEXT:    [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
// CHECK2-NEXT:    [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[VLA7]], i64 [[TMP20]]
// CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[TMP4]], i64 0, i64 1
// CHECK2-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[ARRAYIDX8]], i64 0, i64 0
// CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[ARRAYDECAY]], i64 1
// CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[ARRAYIDX10]], align 4
// CHECK2-NEXT:    [[TMP23:%.*]] = sext i32 [[TMP22]] to i64
// CHECK2-NEXT:    [[LB_ADD_LEN11:%.*]] = add nsw i64 0, [[TMP23]]
// CHECK2-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[TMP4]], i64 0, i64 [[LB_ADD_LEN11]]
// CHECK2-NEXT:    [[ARRAYDECAY13:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[ARRAYIDX12]], i64 0, i64 0
// CHECK2-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDECAY13]], i64 2
// CHECK2-NEXT:    [[TMP24:%.*]] = ptrtoint %struct.S* [[ARRAYIDX14]] to i64
// CHECK2-NEXT:    [[TMP25:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
// CHECK2-NEXT:    [[TMP26:%.*]] = sub i64 [[TMP24]], [[TMP25]]
// CHECK2-NEXT:    [[TMP27:%.*]] = sdiv exact i64 [[TMP26]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP28:%.*]] = add nuw i64 [[TMP27]], 1
// CHECK2-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[VLA15:%.*]] = alloca [[STRUCT_S]], i64 [[TMP28]], align 16
// CHECK2-NEXT:    store i64 [[TMP28]], i64* [[__VLA_EXPR1]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP28]]
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY16:%.*]] = icmp eq %struct.S* [[VLA15]], [[TMP30]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY16]], label [[OMP_ARRAYINIT_DONE21:%.*]], label [[OMP_ARRAYINIT_BODY17:%.*]]
// CHECK2:       omp.arrayinit.body17:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST18:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYINIT_DONE]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT19:%.*]], [[OMP_ARRAYINIT_BODY17]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST18]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST18]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT19]], [[TMP30]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYINIT_DONE21]], label [[OMP_ARRAYINIT_BODY17]]
// CHECK2:       omp.arrayinit.done21:
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast [10 x [4 x %struct.S]]* [[TMP4]] to %struct.S*
// CHECK2-NEXT:    [[TMP32:%.*]] = ptrtoint %struct.S* [[TMP31]] to i64
// CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
// CHECK2-NEXT:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
// CHECK2-NEXT:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP35]]
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[TMP36]] to [10 x [4 x %struct.S]]*
// CHECK2-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP38]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP39]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP40]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP41]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP42]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP43]], [[TMP44]]
// CHECK2-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP45]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP46:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX23:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i64 [[TMP46]]
// CHECK2-NEXT:    [[TMP47:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP47]] to i64
// CHECK2-NEXT:    [[ARRAYIDX24:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX23]], i64 [[IDXPROM]]
// CHECK2-NEXT:    [[TMP48:%.*]] = load i32, i32* [[ARRAYIDX24]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP48]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[ARRAYIDX24]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP49:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP49]], 1
// CHECK2-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP50:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = load i32, i32* [[TMP50]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP51]])
// CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP53:%.*]] = bitcast i32* [[VLA7]] to i8*
// CHECK2-NEXT:    store i8* [[TMP53]], i8** [[TMP52]], align 8
// CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP55:%.*]] = inttoptr i64 [[TMP13]] to i8*
// CHECK2-NEXT:    store i8* [[TMP55]], i8** [[TMP54]], align 8
// CHECK2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP57:%.*]] = bitcast %struct.S* [[VLA15]] to i8*
// CHECK2-NEXT:    store i8* [[TMP57]], i8** [[TMP56]], align 8
// CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP59:%.*]] = inttoptr i64 [[TMP28]] to i8*
// CHECK2-NEXT:    store i8* [[TMP59]], i8** [[TMP58]], align 8
// CHECK2-NEXT:    [[TMP60:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP61:%.*]] = load i32, i32* [[TMP60]], align 4
// CHECK2-NEXT:    [[TMP62:%.*]] = bitcast [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP63:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP61]], i32 2, i64 32, i8* [[TMP62]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP63]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 [[TMP13]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP64]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE30:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[VLA7]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP65:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], align 4
// CHECK2-NEXT:    [[TMP66:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP65]], [[TMP66]]
// CHECK2-NEXT:    store i32 [[ADD27]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE29:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP64]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE29]], label [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done30:
// CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX9]], i64 [[TMP28]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY31:%.*]] = icmp eq %struct.S* [[ARRAYIDX9]], [[TMP67]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY31]], label [[OMP_ARRAYCPY_DONE38:%.*]], label [[OMP_ARRAYCPY_BODY32:%.*]]
// CHECK2:       omp.arraycpy.body32:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST33:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYCPY_DONE30]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT36:%.*]], [[OMP_ARRAYCPY_BODY32]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST34:%.*]] = phi %struct.S* [ [[ARRAYIDX9]], [[OMP_ARRAYCPY_DONE30]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT35:%.*]], [[OMP_ARRAYCPY_BODY32]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST34]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST33]])
// CHECK2-NEXT:    [[TMP68:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST34]] to i8*
// CHECK2-NEXT:    [[TMP69:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP68]], i8* align 4 [[TMP69]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT35]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST34]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT36]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST33]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE37:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT35]], [[TMP67]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE37]], label [[OMP_ARRAYCPY_DONE38]], label [[OMP_ARRAYCPY_BODY32]]
// CHECK2:       omp.arraycpy.done38:
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP61]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 [[TMP13]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY39:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP70]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY39]], label [[OMP_ARRAYCPY_DONE46:%.*]], label [[OMP_ARRAYCPY_BODY40:%.*]]
// CHECK2:       omp.arraycpy.body40:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST41:%.*]] = phi i32* [ [[VLA7]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT44:%.*]], [[OMP_ARRAYCPY_BODY40]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST42:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT43:%.*]], [[OMP_ARRAYCPY_BODY40]] ]
// CHECK2-NEXT:    [[TMP71:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST41]], align 4
// CHECK2-NEXT:    [[TMP72:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST42]], i32 [[TMP71]] monotonic, align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT43]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST42]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT44]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST41]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE45:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT43]], [[TMP70]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE45]], label [[OMP_ARRAYCPY_DONE46]], label [[OMP_ARRAYCPY_BODY40]]
// CHECK2:       omp.arraycpy.done46:
// CHECK2-NEXT:    [[TMP73:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX9]], i64 [[TMP28]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY47:%.*]] = icmp eq %struct.S* [[ARRAYIDX9]], [[TMP73]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY47]], label [[OMP_ARRAYCPY_DONE55:%.*]], label [[OMP_ARRAYCPY_BODY48:%.*]]
// CHECK2:       omp.arraycpy.body48:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST49:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYCPY_DONE46]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT53:%.*]], [[OMP_ARRAYCPY_BODY48]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST50:%.*]] = phi %struct.S* [ [[ARRAYIDX9]], [[OMP_ARRAYCPY_DONE46]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT52:%.*]], [[OMP_ARRAYCPY_BODY48]] ]
// CHECK2-NEXT:    [[TMP74:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP75:%.*]] = load i32, i32* [[TMP74]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP75]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL51:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST50]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST49]])
// CHECK2-NEXT:    [[TMP76:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST50]] to i8*
// CHECK2-NEXT:    [[TMP77:%.*]] = bitcast %struct.S* [[CALL51]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP76]], i8* align 4 [[TMP77]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP75]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT52]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST50]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT53]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST49]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE54:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT52]], [[TMP73]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE54]], label [[OMP_ARRAYCPY_DONE55]], label [[OMP_ARRAYCPY_BODY48]]
// CHECK2:       omp.arraycpy.done55:
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP28]]
// CHECK2-NEXT:    [[ARRAYDESTROY_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA15]], [[TMP78]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_ISEMPTY]], label [[ARRAYDESTROY_DONE56:%.*]], label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP78]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[VLA15]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE56]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done56:
// CHECK2-NEXT:    [[TMP79:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP79]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [4 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [4 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S*
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S*
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = ptrtoint i8* [[TMP22]] to i64
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr i32, i32* [[TMP11]], i64 [[TMP14]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP20]], i64 [[TMP23]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY3:%.*]] = icmp eq %struct.S* [[TMP20]], [[TMP27]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY3]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY4:%.*]]
// CHECK2:       omp.arraycpy.body4:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST5:%.*]] = phi %struct.S* [ [[TMP17]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi %struct.S* [ [[TMP20]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT7:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST6]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST5]])
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]] to i8*
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP28]], i8* align 4 [[TMP29]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT7]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST5]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT7]], [[TMP27]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY4]]
// CHECK2:       omp.arraycpy.done10:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARR:%.*]], [10 x [4 x %struct.S]]* nonnull align 4 dereferenceable(160) [[ARRS:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[ARR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[ARRS_ADDR:%.*]] = alloca [10 x [4 x %struct.S]]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[ARRS4:%.*]] = alloca [10 x [4 x %struct.S]], align 16
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i32* [[ARR]], i32** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    store [10 x [4 x %struct.S]]* [[ARRS]], [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load [10 x [4 x %struct.S]]*, [10 x [4 x %struct.S]]** [[ARRS_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP0]], [[TMP1]]
// CHECK2-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
// CHECK2-NEXT:    [[TMP6:%.*]] = udiv exact i64 [[TMP5]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[VLA3:%.*]] = alloca i32, i64 [[TMP6]], align 16
// CHECK2-NEXT:    store i64 [[TMP6]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr i32, i32* [[VLA3]], i64 [[TMP6]]
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[VLA3]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[VLA3]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS4]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY5:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY5]], label [[OMP_ARRAYINIT_DONE10:%.*]], label [[OMP_ARRAYINIT_BODY6:%.*]]
// CHECK2:       omp.arrayinit.body6:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST7:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[OMP_ARRAYINIT_DONE]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYINIT_BODY6]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST7]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP9]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYINIT_DONE10]], label [[OMP_ARRAYINIT_BODY6]]
// CHECK2:       omp.arrayinit.done10:
// CHECK2-NEXT:    [[LHS_BEGIN:%.*]] = bitcast [10 x [4 x %struct.S]]* [[TMP3]] to %struct.S*
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [10 x [4 x %struct.S]]* [[ARRS4]] to %struct.S*
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA3]], i64 [[TMP18]]
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
// CHECK2-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 [[IDXPROM]]
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[ARRAYIDX12]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP21]], 1
// CHECK2-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i32* [[VLA3]] to i8*
// CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP27:%.*]] = inttoptr i64 [[TMP6]] to i8*
// CHECK2-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP29]], i8** [[TMP28]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP31]], i32 2, i64 24, i8* [[TMP32]], void (i8*, i8*)* @.omp.reduction.reduction_func.4, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP33]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr i32, i32* [[TMP2]], i64 [[TMP6]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP2]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[VLA3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi i32* [ [[TMP2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP35]], [[TMP36]]
// CHECK2-NEXT:    store i32 [[ADD15]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done18:
// CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP37]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE26:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
// CHECK2:       omp.arraycpy.body20:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[OMP_ARRAYCPY_DONE18]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[OMP_ARRAYCPY_DONE18]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST22]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST21]])
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] to i8*
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP38]], i8* align 4 [[TMP39]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT24]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE25:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP37]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_BODY20]]
// CHECK2:       omp.arraycpy.done26:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP31]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr i32, i32* [[TMP2]], i64 [[TMP6]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY27:%.*]] = icmp eq i32* [[TMP2]], [[TMP40]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY27]], label [[OMP_ARRAYCPY_DONE34:%.*]], label [[OMP_ARRAYCPY_BODY28:%.*]]
// CHECK2:       omp.arraycpy.body28:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST29:%.*]] = phi i32* [ [[VLA3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT32:%.*]], [[OMP_ARRAYCPY_BODY28]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST30:%.*]] = phi i32* [ [[TMP2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT31:%.*]], [[OMP_ARRAYCPY_BODY28]] ]
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST29]], align 4
// CHECK2-NEXT:    [[TMP42:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST30]], i32 [[TMP41]] monotonic, align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT31]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST30]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT32]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST29]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE33:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT31]], [[TMP40]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE33]], label [[OMP_ARRAYCPY_DONE34]], label [[OMP_ARRAYCPY_BODY28]]
// CHECK2:       omp.arraycpy.done34:
// CHECK2-NEXT:    [[TMP43:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY35:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP43]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY35]], label [[OMP_ARRAYCPY_DONE43:%.*]], label [[OMP_ARRAYCPY_BODY36:%.*]]
// CHECK2:       omp.arraycpy.body36:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST37:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[OMP_ARRAYCPY_DONE34]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT41:%.*]], [[OMP_ARRAYCPY_BODY36]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST38:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[OMP_ARRAYCPY_DONE34]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT40:%.*]], [[OMP_ARRAYCPY_BODY36]] ]
// CHECK2-NEXT:    [[TMP44:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP45:%.*]] = load i32, i32* [[TMP44]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP45]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL39:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST38]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST37]])
// CHECK2-NEXT:    [[TMP46:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST38]] to i8*
// CHECK2-NEXT:    [[TMP47:%.*]] = bitcast %struct.S* [[CALL39]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP46]], i8* align 4 [[TMP47]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP45]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT40]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST38]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT41]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST37]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE42:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT40]], [[TMP43]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE42]], label [[OMP_ARRAYCPY_DONE43]], label [[OMP_ARRAYCPY_BODY36]]
// CHECK2:       omp.arraycpy.done43:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP31]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN44:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[ARRS4]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN44]], i64 40
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP48]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN44]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE45:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done45:
// CHECK2-NEXT:    [[TMP49:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP49]])
// CHECK2-NEXT:    [[TMP50:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP51:%.*]] = load i32, i32* [[TMP50]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP51]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.4
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S*
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S*
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i32, i32* [[TMP11]], i64 [[TMP14]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP21]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP20]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY3:%.*]] = icmp eq %struct.S* [[TMP20]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY3]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY4:%.*]]
// CHECK2:       omp.arraycpy.body4:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST5:%.*]] = phi %struct.S* [ [[TMP17]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi %struct.S* [ [[TMP20]], [[OMP_ARRAYCPY_DONE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT7:%.*]], [[OMP_ARRAYCPY_BODY4]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST6]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST5]])
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]] to i8*
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP25]], i8* align 4 [[TMP26]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT7]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST5]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT7]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY4]]
// CHECK2:       omp.arraycpy.done10:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARR:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[ARR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[ARR6:%.*]] = alloca [1 x [2 x i32]], align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    store i32* [[ARR]], i32** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP3:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP3]]
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 0
// CHECK2-NEXT:    [[TMP4:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP4]]
// CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX4]], i64 1
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [1 x [2 x i32]], [1 x [2 x i32]]* [[ARR6]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP6:%.*]] = ptrtoint i32* [[TMP2]] to i64
// CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint i32* [[ARRAYIDX3]] to i64
// CHECK2-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP6]], [[TMP7]]
// CHECK2-NEXT:    [[TMP9:%.*]] = sdiv exact i64 [[TMP8]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP10:%.*]] = bitcast [1 x [2 x i32]]* [[ARR6]] to i32*
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[TMP10]], i64 [[TMP9]]
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [2 x i32]]* [[ARR6]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP14]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = mul nsw i64 1, [[TMP1]]
// CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[TMP11]], i64 [[TMP20]]
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
// CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX8]], i64 [[IDXPROM]]
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[ARRAYIDX9]], align 4
// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[INC]], i32* [[ARRAYIDX9]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP23]], 1
// CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
// CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP27]], i8** [[TMP26]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
// CHECK2-NEXT:    [[TMP30:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP29]], i32 1, i64 8, i8* [[TMP30]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP31]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP32]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE15:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT13:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 4
// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
// CHECK2-NEXT:    store i32 [[ADD12]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT13]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE14:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT13]], [[TMP32]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_DONE15]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done15:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr i32, i32* [[ARRAYIDX3]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY16:%.*]] = icmp eq i32* [[ARRAYIDX3]], [[TMP35]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY16]], label [[OMP_ARRAYCPY_DONE23:%.*]], label [[OMP_ARRAYCPY_BODY17:%.*]]
// CHECK2:       omp.arraycpy.body17:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST18:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT21:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST19:%.*]] = phi i32* [ [[ARRAYIDX3]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST18]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST19]], i32 [[TMP36]] monotonic, align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST19]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT21]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST18]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE22:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP35]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_DONE23]], label [[OMP_ARRAYCPY_BODY17]]
// CHECK2:       omp.arraycpy.done23:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP38]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP39]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[_TMP4:%.*]] = alloca %struct.S**, align 8
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 0
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 4
// CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
// CHECK2-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
// CHECK2-NEXT:    [[TMP6:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK2-NEXT:    [[TMP7:%.*]] = sub i64 [[TMP5]], [[TMP6]]
// CHECK2-NEXT:    [[TMP8:%.*]] = sdiv exact i64 [[TMP7]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP9:%.*]] = add nuw i64 [[TMP8]], 1
// CHECK2-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP11:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP11]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP9]], align 16
// CHECK2-NEXT:    store i64 [[TMP9]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP9]]
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP13:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = load %struct.S*, %struct.S** [[TMP13]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[TMP14]] to i64
// CHECK2-NEXT:    [[TMP16:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK2-NEXT:    [[TMP17:%.*]] = sub i64 [[TMP15]], [[TMP16]]
// CHECK2-NEXT:    [[TMP18:%.*]] = sdiv exact i64 [[TMP17]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP18]]
// CHECK2-NEXT:    store %struct.S** [[_TMP5]], %struct.S*** [[_TMP4]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP19]], %struct.S** [[_TMP5]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP23]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[VLA]] to i8*
// CHECK2-NEXT:    store i8* [[TMP32]], i8** [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP9]] to i8*
// CHECK2-NEXT:    store i8* [[TMP34]], i8** [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], i32 1, i64 16, i8* [[TMP37]], void (i8*, i8*)* @.omp.reduction.reduction_func.8, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP38]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 [[TMP9]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP39]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE11:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST8:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT9:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST8]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP40:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]] to i8*
// CHECK2-NEXT:    [[TMP41:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT9]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE10:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT9]], [[TMP39]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done11:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 [[TMP9]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY12:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP42]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY12]], label [[OMP_ARRAYCPY_DONE20:%.*]], label [[OMP_ARRAYCPY_BODY13:%.*]]
// CHECK2:       omp.arraycpy.body13:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST14:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[TMP43]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL16:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST15]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST14]])
// CHECK2-NEXT:    [[TMP45:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]] to i8*
// CHECK2-NEXT:    [[TMP46:%.*]] = bitcast %struct.S* [[CALL16]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP45]], i8* align 4 [[TMP46]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST14]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE19:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP42]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_BODY13]]
// CHECK2:       omp.arraycpy.done20:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP9]]
// CHECK2-NEXT:    [[ARRAYDESTROY_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP47]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_ISEMPTY]], label [[ARRAYDESTROY_DONE21:%.*]], label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP47]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[VLA]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE21]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done21:
// CHECK2-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK2-NEXT:    [[TMP49:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = load i32, i32* [[TMP49]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP50]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.8
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 [[TMP14]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP16:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP16]], i8* align 4 [[TMP17]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR24:%.*]] = alloca [1 x [6 x %struct.S]], align 16
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca %struct.S**, align 8
// CHECK2-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 1
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 1
// CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
// CHECK2-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.10, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE12:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST9:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT10:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST9]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT10]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE11:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT10]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_DONE12]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done12:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY13:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY13]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY14:%.*]]
// CHECK2:       omp.arraycpy.body14:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST15:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL17:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST16]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST15]])
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL17]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST15]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT18]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY14]]
// CHECK2:       omp.arraycpy.done21:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN22:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN22]], i64 6
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN22]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE23:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done23:
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.10
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR24:%.*]] = alloca [1 x [6 x %struct.S]], align 16
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca %struct.S**, align 8
// CHECK2-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 1
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 1
// CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
// CHECK2-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.12, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE12:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST9:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT10:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST9]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT10]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST9]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE11:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT10]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_DONE12]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done12:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX1]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY13:%.*]] = icmp eq %struct.S* [[ARRAYIDX1]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY13]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY14:%.*]]
// CHECK2:       omp.arraycpy.body14:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST15:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi %struct.S* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY14]] ]
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL17:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST16]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST15]])
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL17]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST15]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT18]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY14]]
// CHECK2:       omp.arraycpy.done21:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN22:%.*]] = getelementptr inbounds [1 x [6 x %struct.S]], [1 x [6 x %struct.S]]* [[VAR24]], i32 0, i32 0, i32 0
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN22]], i64 6
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN22]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE23:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done23:
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.12
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 6
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..13
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.S*** nonnull align 8 dereferenceable(8) [[VAR2:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR2_ADDR:%.*]] = alloca %struct.S***, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR24:%.*]] = alloca [[STRUCT_S:%.*]], align 4
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca %struct.S**, align 8
// CHECK2-NEXT:    [[_TMP6:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store %struct.S*** [[VAR2]], %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S***, %struct.S**** [[VAR2_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP1]], i64 1
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX]], align 8
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP2]], i64 1
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 1
// CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 1
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
// CHECK2-NEXT:    [[TMP5:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S*, %struct.S** [[TMP5]], align 8
// CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VAR24]], i64 [[TMP10]]
// CHECK2-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
// CHECK2-NEXT:    store %struct.S* [[TMP11]], %struct.S** [[_TMP6]], align 8
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP14]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
// CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
// CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]])
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast %struct.S* [[VAR24]] to i8*
// CHECK2-NEXT:    store i8* [[TMP22]], i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], i32 1, i64 8, i8* [[TMP23]], void (i8*, i8*)* @.omp.reduction.reduction_func.14, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYIDX1]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast %struct.S* [[ARRAYIDX1]] to i8*
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP25]], i8* align 4 [[TMP26]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYIDX1]], %struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[ARRAYIDX1]] to i8*
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[CALL9]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP27]], i8* align 4 [[TMP28]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR24]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP13]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.14
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[TMP11]], %struct.S* nonnull align 4 dereferenceable(4) [[TMP8]])
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast %struct.S* [[TMP11]] to i8*
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP12]], i8* align 4 [[TMP13]], i64 4, i1 false)
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [5 x %struct.S]* nonnull align 4 dereferenceable(20) [[VVAR2:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VVAR2_ADDR:%.*]] = alloca [5 x %struct.S]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VVAR22:%.*]] = alloca [5 x %struct.S], align 16
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store [5 x %struct.S]* [[VVAR2]], [5 x %struct.S]** [[VVAR2_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load [5 x %struct.S]*, [5 x %struct.S]** [[VVAR2_ADDR]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[TMP0]], i64 0, i64 0
// CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[TMP0]], i64 0, i64 4
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR22]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 5
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP1]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP2:%.*]] = bitcast [5 x %struct.S]* [[TMP0]] to %struct.S*
// CHECK2-NEXT:    [[TMP3:%.*]] = ptrtoint %struct.S* [[TMP2]] to i64
// CHECK2-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
// CHECK2-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP7:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP7]], i64 [[TMP6]]
// CHECK2-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[TMP8]] to [5 x %struct.S]*
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
// CHECK2-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK2-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP22]], i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    [[TMP25:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP24]], i32 1, i64 8, i8* [[TMP25]], void (i8*, i8*)* @.omp.reduction.reduction_func.16, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP26]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 5
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP27]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE8:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST5:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT6:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST5]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST5]] to i8*
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP28]], i8* align 4 [[TMP29]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT6]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST5]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE7:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT6]], [[TMP27]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE7]], label [[OMP_ARRAYCPY_DONE8]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done8:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP24]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 5
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY9:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP30]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY9]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY10:%.*]]
// CHECK2:       omp.arraycpy.body10:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST11:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY10]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT14:%.*]], [[OMP_ARRAYCPY_BODY10]] ]
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP31]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP32]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL13:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST12]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST11]])
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST12]] to i8*
// CHECK2-NEXT:    [[TMP34:%.*]] = bitcast %struct.S* [[CALL13]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP33]], i8* align 4 [[TMP34]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP32]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT14]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT15]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST11]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT14]], [[TMP30]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY10]]
// CHECK2:       omp.arraycpy.done17:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP24]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN18:%.*]] = getelementptr inbounds [5 x %struct.S], [5 x %struct.S]* [[VVAR22]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN18]], i64 5
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP35]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN18]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE19:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done19:
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP36]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP37]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.16
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 5
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..17
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR34:%.*]] = alloca [2 x %struct.S], align 4
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 2
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP4]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
// CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.18, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE11:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST8:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT9:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST8]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT9]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE10:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT9]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done11:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY12:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY12]], label [[OMP_ARRAYCPY_DONE20:%.*]], label [[OMP_ARRAYCPY_BODY13:%.*]]
// CHECK2:       omp.arraycpy.body13:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST14:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL16:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST15]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST14]])
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL16]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST14]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE19:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_BODY13]]
// CHECK2:       omp.arraycpy.done20:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN21:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN21]], i64 2
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN21]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE22:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done22:
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.18
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..19
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR34:%.*]] = alloca [2 x %struct.S], align 4
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP4]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
// CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP16]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP18]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
// CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP21]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP24]])
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func.20, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE11:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST8:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT9:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST8]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT9]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST8]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE10:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT9]], [[TMP31]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_DONE11]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done11:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY12:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY12]], label [[OMP_ARRAYCPY_DONE20:%.*]], label [[OMP_ARRAYCPY_BODY13:%.*]]
// CHECK2:       omp.arraycpy.body13:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST14:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT18:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY13]] ]
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL16:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST15]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST14]])
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S* [[CALL16]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP37]], i8* align 4 [[TMP38]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT18]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST14]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE19:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP34]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_BODY13]]
// CHECK2:       omp.arraycpy.done20:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN21:%.*]] = getelementptr inbounds [2 x %struct.S], [2 x %struct.S]* [[VAR34]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN21]], i64 2
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP39]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN21]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE22:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done22:
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[TMP40]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP41]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.20
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 2
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..21
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
// CHECK2-NEXT:    [[_TMP4:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
// CHECK2-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP6:%.*]] = sub i64 [[TMP4]], [[TMP5]]
// CHECK2-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S:%.*]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP8:%.*]] = add nuw i64 [[TMP7]], 1
// CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP10:%.*]] = call i8* @llvm.stacksave()
// CHECK2-NEXT:    store i8* [[TMP10]], i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP8]], align 16
// CHECK2-NEXT:    store i64 [[TMP8]], i64* [[__VLA_EXPR0]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP8]]
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP12:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast [4 x %struct.S]* [[TMP12]] to %struct.S*
// CHECK2-NEXT:    [[TMP14:%.*]] = ptrtoint %struct.S* [[TMP13]] to i64
// CHECK2-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
// CHECK2-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP17]]
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast %struct.S* [[TMP18]] to [4 x %struct.S]*
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP19]], [4 x %struct.S]** [[_TMP4]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP23]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP24]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP25]], [[TMP26]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP27]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP28]], 1
// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
// CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S* [[VLA]] to i8*
// CHECK2-NEXT:    store i8* [[TMP32]], i8** [[TMP31]], align 8
// CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP34:%.*]] = inttoptr i64 [[TMP8]] to i8*
// CHECK2-NEXT:    store i8* [[TMP34]], i8** [[TMP33]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP35]], align 4
// CHECK2-NEXT:    [[TMP37:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP38:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], i32 1, i64 16, i8* [[TMP37]], void (i8*, i8*)* @.omp.reduction.reduction_func.22, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP38]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 [[TMP8]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP39]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST7:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST7]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP40:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]] to i8*
// CHECK2-NEXT:    [[TMP41:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP39]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done10:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[ARRAYIDX]], i64 [[TMP8]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq %struct.S* [[ARRAYIDX]], [[TMP42]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
// CHECK2:       omp.arraycpy.body12:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi %struct.S* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi %struct.S* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[TMP43]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL15:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST14]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST13]])
// CHECK2-NEXT:    [[TMP45:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]] to i8*
// CHECK2-NEXT:    [[TMP46:%.*]] = bitcast %struct.S* [[CALL15]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP45]], i8* align 4 [[TMP46]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP44]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP42]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY12]]
// CHECK2:       omp.arraycpy.done19:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP36]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP8]]
// CHECK2-NEXT:    [[ARRAYDESTROY_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP47]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_ISEMPTY]], label [[ARRAYDESTROY_DONE20:%.*]], label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP47]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[VLA]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE20]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done20:
// CHECK2-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
// CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
// CHECK2-NEXT:    [[TMP49:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP50:%.*]] = load i32, i32* [[TMP49]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP50]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.22
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = ptrtoint i8* [[TMP13]] to i64
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 [[TMP14]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP16:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP16]], i8* align 4 [[TMP17]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..23
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [4 x %struct.S]* nonnull align 4 dereferenceable(16) [[VAR3:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR3_ADDR:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR33:%.*]] = alloca [4 x %struct.S], align 16
// CHECK2-NEXT:    [[_TMP4:%.*]] = alloca [4 x %struct.S]*, align 8
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[VAR3]], [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[VAR3_ADDR]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP0]], [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[TMP]], align 8
// CHECK2-NEXT:    store [4 x %struct.S]* [[TMP1]], [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[VAR33]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[ARRAY_BEGIN]], i64 4
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[ARRAY_BEGIN]], [[TMP3]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    store [4 x %struct.S]* [[VAR33]], [4 x %struct.S]** [[_TMP4]], align 8
// CHECK2-NEXT:    [[LHS_BEGIN:%.*]] = bitcast [4 x %struct.S]* [[TMP2]] to %struct.S*
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [4 x %struct.S]* [[VAR33]] to %struct.S*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP14]])
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP16:%.*]] = bitcast %struct.S* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 1, i64 8, i8* [[TMP19]], void (i8*, i8*)* @.omp.reduction.reduction_func.24, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP20]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP21]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST7:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST7]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]] to i8*
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP22]], i8* align 4 [[TMP23]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST7]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP21]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done10:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[LHS_BEGIN]], i64 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq %struct.S* [[LHS_BEGIN]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
// CHECK2:       omp.arraycpy.body12:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi %struct.S* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi %struct.S* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP26]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL15:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST14]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST13]])
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]] to i8*
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S* [[CALL15]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP27]], i8* align 4 [[TMP28]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP26]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT17]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP24]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY12]]
// CHECK2:       omp.arraycpy.done19:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN20:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[VAR33]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAY_BEGIN20]], i64 4
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP29]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIfED1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN20]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE21:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done21:
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP31]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.24
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S:%.*]], %struct.S* [[TMP11]], i64 4
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S* @_ZN1SIfEanERKS0_(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], %struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done2:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiLi42EET_v
// CHECK2-SAME: () #[[ATTR7]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[T:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK2-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
// CHECK2-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
// CHECK2-NEXT:    [[VAR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[VAR1:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK2-NEXT:    [[ARR:%.*]] = alloca [42 x %struct.S.0], align 16
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]])
// CHECK2-NEXT:    store i32 0, i32* [[T_VAR]], align 4
// CHECK2-NEXT:    [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiLi42EET_v.vec to i8*), i64 8, i1 false)
// CHECK2-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i64 0, i64 0
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 1)
// CHECK2-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYINIT_BEGIN]], i64 1
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 2)
// CHECK2-NEXT:    store %struct.S.0* [[TEST]], %struct.S.0** [[VAR]], align 8
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR1]])
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[ARR]], i32 0, i32 0
// CHECK2-NEXT:    [[ARRAYCTOR_END:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 42
// CHECK2-NEXT:    br label [[ARRAYCTOR_LOOP:%.*]]
// CHECK2:       arrayctor.loop:
// CHECK2-NEXT:    [[ARRAYCTOR_CUR:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[ARRAYCTOR_NEXT:%.*]], [[ARRAYCTOR_LOOP]] ]
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYCTOR_CUR]])
// CHECK2-NEXT:    [[ARRAYCTOR_NEXT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYCTOR_CUR]], i64 1
// CHECK2-NEXT:    [[ARRAYCTOR_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYCTOR_NEXT]], [[ARRAYCTOR_END]]
// CHECK2-NEXT:    br i1 [[ARRAYCTOR_DONE]], label [[ARRAYCTOR_CONT:%.*]], label [[ARRAYCTOR_LOOP]]
// CHECK2:       arrayctor.cont:
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.S.0*, %struct.S.0*, i32*, [2 x i32]*, [2 x %struct.S.0]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), i32* [[T_VAR]], %struct.S.0* [[TMP1]], %struct.S.0* [[VAR1]], i32* [[T_VAR1]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]])
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [2 x i32]*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..27 to void (i32*, i32*, ...)*), i32* [[T_VAR]], [2 x i32]* [[VEC]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP2]])
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR]], align 8
// CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [42 x %struct.S.0]*, [2 x i32]*, i32*, [2 x %struct.S.0]*, %struct.S.0*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [42 x %struct.S.0]* [[ARR]], [2 x i32]* [[VEC]], i32* [[T_VAR]], [2 x %struct.S.0]* [[S_ARR]], %struct.S.0* [[TMP3]])
// CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK2-NEXT:    [[ARRAY_BEGIN1:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[ARR]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN1]], i64 42
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP4]], [[ARRAYCTOR_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN1]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done2:
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR1]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAY_BEGIN3:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[S_ARR]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN3]], i64 2
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY4:%.*]]
// CHECK2:       arraydestroy.body4:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST5:%.*]] = phi %struct.S.0* [ [[TMP5]], [[ARRAYDESTROY_DONE2]] ], [ [[ARRAYDESTROY_ELEMENT6:%.*]], [[ARRAYDESTROY_BODY4]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT6]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST5]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT6]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE7:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT6]], [[ARRAY_BEGIN3]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE7]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY4]]
// CHECK2:       arraydestroy.done8:
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR5]]
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[RETVAL]], align 4
// CHECK2-NEXT:    ret i32 [[TMP6]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP0:%.*]] = load volatile double, double* @g, align 8
// CHECK2-NEXT:    [[CONV:%.*]] = fptrunc double [[TMP0]] to float
// CHECK2-NEXT:    store float [[CONV]], float* [[F]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]], float [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store float [[A]], float* [[A_ADDR]], align 4
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP0:%.*]] = load float, float* [[A_ADDR]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = fpext float [[TMP0]] to double
// CHECK2-NEXT:    [[TMP1:%.*]] = load volatile double, double* @g, align 8
// CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], [[TMP1]]
// CHECK2-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD]] to float
// CHECK2-NEXT:    store float [[CONV2]], float* [[F]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
// CHECK2-SAME: (%struct.S* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S*, align 8
// CHECK2-NEXT:    store %struct.S* [[THIS]], %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S*, %struct.S** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void @_ZN1SIiEC2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK2-NEXT:    call void @_ZN1SIiEC2Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]], i32 [[TMP0]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..25
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR1:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR1:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[VAR1_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[T_VAR1_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK2-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK2-NEXT:    [[_TMP5:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[VAR16:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK2-NEXT:    [[T_VAR17:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [4 x i8*], align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK2-NEXT:    [[REF_TMP22:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[VAR1]], %struct.S.0** [[VAR1_ADDR]], align 8
// CHECK2-NEXT:    store i32* [[T_VAR1]], i32** [[T_VAR1_ADDR]], align 8
// CHECK2-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR1_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[T_VAR1_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP1]], %struct.S.0** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP6]], %struct.S.0** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK2-NEXT:    store %struct.S.0* [[VAR4]], %struct.S.0** [[_TMP5]], align 8
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK2-NEXT:    store i32 2147483647, i32* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP10]], 1
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
// CHECK2-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP4]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 [[TMP16]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP18:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP5]], align 8
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP19]] to i64
// CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[TMP5]], i64 0, i64 [[IDXPROM9]]
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast %struct.S.0* [[ARRAYIDX10]] to i8*
// CHECK2-NEXT:    [[TMP21:%.*]] = bitcast %struct.S.0* [[TMP18]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP20]], i8* align 4 [[TMP21]], i64 4, i1 false)
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
// CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP24:%.*]] = bitcast i32* [[T_VAR3]] to i8*
// CHECK2-NEXT:    store i8* [[TMP24]], i8** [[TMP23]], align 8
// CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast %struct.S.0* [[VAR4]] to i8*
// CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S.0* [[VAR16]] to i8*
// CHECK2-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP30:%.*]] = bitcast i32* [[T_VAR17]] to i8*
// CHECK2-NEXT:    store i8* [[TMP30]], i8** [[TMP29]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = bitcast [4 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP32:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], i32 4, i64 32, i8* [[TMP31]], void (i8*, i8*)* @.omp.reduction.reduction_func.26, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP32]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP33]], [[TMP34]]
// CHECK2-NEXT:    store i32 [[ADD12]], i32* [[TMP0]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S.0* @_ZN1SIiEanERKS0_(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast %struct.S.0* [[TMP7]] to i8*
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast %struct.S.0* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK2-NEXT:    [[CALL13:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[CALL13]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK2:       land.rhs:
// CHECK2-NEXT:    [[CALL14:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK2-NEXT:    [[TOBOOL15:%.*]] = icmp ne i32 [[CALL14]], 0
// CHECK2-NEXT:    br label [[LAND_END]]
// CHECK2:       land.end:
// CHECK2-NEXT:    [[TMP37:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL15]], [[LAND_RHS]] ]
// CHECK2-NEXT:    [[CONV:%.*]] = zext i1 [[TMP37]] to i32
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[CONV]])
// CHECK2-NEXT:    [[TMP38:%.*]] = bitcast %struct.S.0* [[TMP2]] to i8*
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP38]], i8* align 4 [[TMP39]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    [[TMP41:%.*]] = load i32, i32* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[CMP16:%.*]] = icmp slt i32 [[TMP40]], [[TMP41]]
// CHECK2-NEXT:    br i1 [[CMP16]], label [[COND_TRUE17:%.*]], label [[COND_FALSE18:%.*]]
// CHECK2:       cond.true17:
// CHECK2-NEXT:    [[TMP42:%.*]] = load i32, i32* [[TMP3]], align 4
// CHECK2-NEXT:    br label [[COND_END19:%.*]]
// CHECK2:       cond.false18:
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[T_VAR17]], align 4
// CHECK2-NEXT:    br label [[COND_END19]]
// CHECK2:       cond.end19:
// CHECK2-NEXT:    [[COND20:%.*]] = phi i32 [ [[TMP42]], [[COND_TRUE17]] ], [ [[TMP43]], [[COND_FALSE18]] ]
// CHECK2-NEXT:    store i32 [[COND20]], i32* [[TMP3]], align 4
// CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP45:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP44]] monotonic, align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL21:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S.0* @_ZN1SIiEanERKS0_(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP7]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]])
// CHECK2-NEXT:    [[TMP46:%.*]] = bitcast %struct.S.0* [[TMP7]] to i8*
// CHECK2-NEXT:    [[TMP47:%.*]] = bitcast %struct.S.0* [[CALL21]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP46]], i8* align 4 [[TMP47]], i64 4, i1 false)
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL23:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP2]])
// CHECK2-NEXT:    [[TOBOOL24:%.*]] = icmp ne i32 [[CALL23]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL24]], label [[LAND_RHS25:%.*]], label [[LAND_END28:%.*]]
// CHECK2:       land.rhs25:
// CHECK2-NEXT:    [[CALL26:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]])
// CHECK2-NEXT:    [[TOBOOL27:%.*]] = icmp ne i32 [[CALL26]], 0
// CHECK2-NEXT:    br label [[LAND_END28]]
// CHECK2:       land.end28:
// CHECK2-NEXT:    [[TMP48:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE2]] ], [ [[TOBOOL27]], [[LAND_RHS25]] ]
// CHECK2-NEXT:    [[CONV29:%.*]] = zext i1 [[TMP48]] to i32
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP22]], i32 [[CONV29]])
// CHECK2-NEXT:    [[TMP49:%.*]] = bitcast %struct.S.0* [[TMP2]] to i8*
// CHECK2-NEXT:    [[TMP50:%.*]] = bitcast %struct.S.0* [[REF_TMP22]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP49]], i8* align 4 [[TMP50]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP22]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP9]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[TMP51:%.*]] = load i32, i32* [[T_VAR17]], align 4
// CHECK2-NEXT:    [[TMP52:%.*]] = atomicrmw min i32* [[TMP3]], i32 [[TMP51]] monotonic, align 4
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR16]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR5]]
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.26
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [4 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [4 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
// CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8* [[TMP19]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 2
// CHECK2-NEXT:    [[TMP22:%.*]] = load i8*, i8** [[TMP21]], align 8
// CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8* [[TMP22]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP5]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i8* [[TMP25]] to i32*
// CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[TMP3]], i64 0, i64 3
// CHECK2-NEXT:    [[TMP28:%.*]] = load i8*, i8** [[TMP27]], align 8
// CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8* [[TMP28]] to i32*
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
// CHECK2-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(4) %struct.S.0* @_ZN1SIiEanERKS0_(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP17]], %struct.S.0* nonnull align 4 dereferenceable(4) [[TMP14]])
// CHECK2-NEXT:    [[TMP32:%.*]] = bitcast %struct.S.0* [[TMP17]] to i8*
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S.0* [[CALL]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP32]], i8* align 4 [[TMP33]], i64 4, i1 false)
// CHECK2-NEXT:    [[CALL2:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP23]])
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[CALL2]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK2:       land.rhs:
// CHECK2-NEXT:    [[CALL3:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[TMP20]])
// CHECK2-NEXT:    [[TOBOOL4:%.*]] = icmp ne i32 [[CALL3]], 0
// CHECK2-NEXT:    br label [[LAND_END]]
// CHECK2:       land.end:
// CHECK2-NEXT:    [[TMP34:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL4]], [[LAND_RHS]] ]
// CHECK2-NEXT:    [[CONV:%.*]] = zext i1 [[TMP34]] to i32
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[CONV]])
// CHECK2-NEXT:    [[TMP35:%.*]] = bitcast %struct.S.0* [[TMP23]] to i8*
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP35]], i8* align 4 [[TMP36]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP26]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP37]], [[TMP38]]
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP29]], align 4
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[TMP26]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP39]], [[COND_TRUE]] ], [ [[TMP40]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[TMP29]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiEanERKS0_
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[TMP0:%.*]]) #[[ATTR7]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP0]], %struct.S.0** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    ret %struct.S.0* [[THIS1]]
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiEcviEv
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) #[[ATTR7]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    ret i32 0
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    call void @_ZN1SIiED2Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR5]]
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..27
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK2-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK2-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP10:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP3]], %struct.S.0** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP14]] to i64
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 [[TMP13]], i32* [[ARRAYIDX]], align 4
// CHECK2-NEXT:    [[TMP15:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP16]] to i64
// CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[TMP2]], i64 0, i64 [[IDXPROM5]]
// CHECK2-NEXT:    [[TMP17:%.*]] = bitcast %struct.S.0* [[ARRAYIDX6]] to i8*
// CHECK2-NEXT:    [[TMP18:%.*]] = bitcast %struct.S.0* [[TMP15]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP17]], i8* align 4 [[TMP18]], i64 4, i1 false)
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
// CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]])
// CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i32* [[T_VAR3]] to i8*
// CHECK2-NEXT:    store i8* [[TMP21]], i8** [[TMP20]], align 8
// CHECK2-NEXT:    [[TMP22:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP23:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 1, i64 8, i8* [[TMP22]], void (i8*, i8*)* @.omp.reduction.reduction_func.28, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP23]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP0]], align 4
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP24]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK2:       land.rhs:
// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TOBOOL8:%.*]] = icmp ne i32 [[TMP25]], 0
// CHECK2-NEXT:    br label [[LAND_END]]
// CHECK2:       land.end:
// CHECK2-NEXT:    [[TMP26:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL8]], [[LAND_RHS]] ]
// CHECK2-NEXT:    [[CONV:%.*]] = zext i1 [[TMP26]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[TMP0]], align 4
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TOBOOL9:%.*]] = icmp ne i32 [[TMP27]], 0
// CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, i32* [[TMP0]] monotonic, align 4
// CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
// CHECK2:       atomic_cont:
// CHECK2-NEXT:    [[TMP28:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP34:%.*]], [[LAND_END14:%.*]] ]
// CHECK2-NEXT:    store i32 [[TMP28]], i32* [[_TMP10]], align 4
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[_TMP10]], align 4
// CHECK2-NEXT:    [[TOBOOL11:%.*]] = icmp ne i32 [[TMP29]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL11]], label [[LAND_RHS12:%.*]], label [[LAND_END14]]
// CHECK2:       land.rhs12:
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[T_VAR3]], align 4
// CHECK2-NEXT:    [[TOBOOL13:%.*]] = icmp ne i32 [[TMP30]], 0
// CHECK2-NEXT:    br label [[LAND_END14]]
// CHECK2:       land.end14:
// CHECK2-NEXT:    [[TMP31:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL13]], [[LAND_RHS12]] ]
// CHECK2-NEXT:    [[CONV15:%.*]] = zext i1 [[TMP31]] to i32
// CHECK2-NEXT:    store i32 [[CONV15]], i32* [[ATOMIC_TEMP]], align 4
// CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[ATOMIC_TEMP]], align 4
// CHECK2-NEXT:    [[TMP33:%.*]] = cmpxchg i32* [[TMP0]], i32 [[TMP28]], i32 [[TMP32]] monotonic monotonic, align 4
// CHECK2-NEXT:    [[TMP34]] = extractvalue { i32, i1 } [[TMP33]], 0
// CHECK2-NEXT:    [[TMP35:%.*]] = extractvalue { i32, i1 } [[TMP33]], 1
// CHECK2-NEXT:    br i1 [[TMP35]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
// CHECK2:       atomic_exit:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP6]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.28
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
// CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
// CHECK2-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
// CHECK2:       land.rhs:
// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
// CHECK2-NEXT:    [[TOBOOL2:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK2-NEXT:    br label [[LAND_END]]
// CHECK2:       land.end:
// CHECK2-NEXT:    [[TMP14:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
// CHECK2-NEXT:    [[CONV:%.*]] = zext i1 [[TMP14]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[TMP11]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..29
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [42 x %struct.S.0]* nonnull align 4 dereferenceable(168) [[ARR:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], [2 x %struct.S.0]* nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], %struct.S.0* nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[ARR_ADDR:%.*]] = alloca [42 x %struct.S.0]*, align 8
// CHECK2-NEXT:    [[VEC_ADDR:%.*]] = alloca [2 x i32]*, align 8
// CHECK2-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT:    [[S_ARR_ADDR:%.*]] = alloca [2 x %struct.S.0]*, align 8
// CHECK2-NEXT:    [[VAR_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[TMP:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[ARR4:%.*]] = alloca [40 x %struct.S.0], align 16
// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK2-NEXT:    [[REF_TMP20:%.*]] = alloca [[STRUCT_S_0]], align 4
// CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT:    store [42 x %struct.S.0]* [[ARR]], [42 x %struct.S.0]** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    store [2 x i32]* [[VEC]], [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    store [2 x %struct.S.0]* [[S_ARR]], [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[VAR]], %struct.S.0** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP0:%.*]] = load [42 x %struct.S.0]*, [42 x %struct.S.0]** [[ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP1:%.*]] = load [2 x i32]*, [2 x i32]** [[VEC_ADDR]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = load [2 x %struct.S.0]*, [2 x %struct.S.0]** [[S_ARR_ADDR]], align 8
// CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S.0*, %struct.S.0** [[VAR_ADDR]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP4]], %struct.S.0** [[TMP]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = load %struct.S.0*, %struct.S.0** [[TMP]], align 8
// CHECK2-NEXT:    store %struct.S.0* [[TMP5]], %struct.S.0** [[_TMP1]], align 8
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[TMP0]], i64 0, i64 1
// CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [42 x %struct.S.0], [42 x %struct.S.0]* [[TMP0]], i64 0, i64 40
// CHECK2-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [40 x %struct.S.0], [40 x %struct.S.0]* [[ARR4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAY_BEGIN]], [[TMP6]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
// CHECK2:       omp.arrayinit.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
// CHECK2:       omp.arrayinit.done:
// CHECK2-NEXT:    [[TMP7:%.*]] = bitcast [42 x %struct.S.0]* [[TMP0]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S.0* [[TMP7]] to i64
// CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S.0* [[ARRAYIDX]] to i64
// CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S.0* getelementptr ([[STRUCT_S_0]], %struct.S.0* null, i32 1) to i64)
// CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[TMP12]], i64 [[TMP11]]
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP13]] to [42 x %struct.S.0]*
// CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP17]], 1
// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2:       cond.true:
// CHECK2-NEXT:    br label [[COND_END:%.*]]
// CHECK2:       cond.false:
// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    br label [[COND_END]]
// CHECK2:       cond.end:
// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP18]], [[COND_FALSE]] ]
// CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT:    store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2:       omp.inner.for.cond:
// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
// CHECK2-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
// CHECK2:       omp.inner.for.cond.cleanup:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
// CHECK2:       omp.inner.for.body:
// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP22]], 1
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP24]] to i64
// CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
// CHECK2-NEXT:    store i32 [[TMP23]], i32* [[ARRAYIDX6]], align 4
// CHECK2-NEXT:    [[TMP25:%.*]] = load %struct.S.0*, %struct.S.0** [[_TMP1]], align 8
// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
// CHECK2-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP26]] to i64
// CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S.0], [2 x %struct.S.0]* [[TMP3]], i64 0, i64 [[IDXPROM7]]
// CHECK2-NEXT:    [[TMP27:%.*]] = bitcast %struct.S.0* [[ARRAYIDX8]] to i8*
// CHECK2-NEXT:    [[TMP28:%.*]] = bitcast %struct.S.0* [[TMP25]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP27]], i8* align 4 [[TMP28]], i64 4, i1 false)
// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2:       omp.body.continue:
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2:       omp.inner.for.inc:
// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP29]], 1
// CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK2:       omp.inner.for.end:
// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2:       omp.loop.exit:
// CHECK2-NEXT:    [[TMP30:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP30]], align 4
// CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP31]])
// CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP33:%.*]] = bitcast %struct.S.0* [[RHS_BEGIN]] to i8*
// CHECK2-NEXT:    store i8* [[TMP33]], i8** [[TMP32]], align 8
// CHECK2-NEXT:    [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
// CHECK2-NEXT:    [[TMP36:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK2-NEXT:    [[TMP37:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP35]], i32 1, i64 8, i8* [[TMP36]], void (i8*, i8*)* @.omp.reduction.reduction_func.30, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    switch i32 [[TMP37]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK2-NEXT:    ]
// CHECK2:       .omp.reduction.case1:
// CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAYIDX]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[ARRAYIDX]], [[TMP38]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE15:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi %struct.S.0* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT13:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST10]])
// CHECK2-NEXT:    [[CALL11:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[CALL]], [[CALL11]]
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[ADD12]])
// CHECK2-NEXT:    [[TMP39:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST10]] to i8*
// CHECK2-NEXT:    [[TMP40:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP39]], i8* align 4 [[TMP40]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT13]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE14:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT13]], [[TMP38]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_DONE15]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done15:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP35]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.case2:
// CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[ARRAYIDX]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY16:%.*]] = icmp eq %struct.S.0* [[ARRAYIDX]], [[TMP41]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY16]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY17:%.*]]
// CHECK2:       omp.arraycpy.body17:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST18:%.*]] = phi %struct.S.0* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST19:%.*]] = phi %struct.S.0* [ [[ARRAYIDX]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY17]] ]
// CHECK2-NEXT:    [[TMP42:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[TMP42]], align 4
// CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP43]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[CALL21:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST19]])
// CHECK2-NEXT:    [[CALL22:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST18]])
// CHECK2-NEXT:    [[ADD23:%.*]] = add nsw i32 [[CALL21]], [[CALL22]]
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP20]], i32 [[ADD23]])
// CHECK2-NEXT:    [[TMP44:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST19]] to i8*
// CHECK2-NEXT:    [[TMP45:%.*]] = bitcast %struct.S.0* [[REF_TMP20]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP44]], i8* align 4 [[TMP45]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP20]]) #[[ATTR5]]
// CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP43]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST19]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST18]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP41]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY17]]
// CHECK2:       omp.arraycpy.done27:
// CHECK2-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP35]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK2:       .omp.reduction.default:
// CHECK2-NEXT:    [[ARRAY_BEGIN28:%.*]] = getelementptr inbounds [40 x %struct.S.0], [40 x %struct.S.0]* [[ARR4]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAY_BEGIN28]], i64 40
// CHECK2-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
// CHECK2:       arraydestroy.body:
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP46]], [[DOTOMP_REDUCTION_DEFAULT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
// CHECK2-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], %struct.S.0* [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR5]]
// CHECK2-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq %struct.S.0* [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN28]]
// CHECK2-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE29:%.*]], label [[ARRAYDESTROY_BODY]]
// CHECK2:       arraydestroy.done29:
// CHECK2-NEXT:    [[TMP47:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT:    [[TMP48:%.*]] = load i32, i32* [[TMP47]], align 4
// CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP48]])
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.30
// CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR6]] {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK2-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
// CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
// CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.S.0*
// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[TMP11]], i64 40
// CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq %struct.S.0* [[TMP11]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
// CHECK2:       omp.arraycpy.body:
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S.0* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
// CHECK2-NEXT:    [[CALL:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
// CHECK2-NEXT:    [[CALL2:%.*]] = call i32 @_ZN1SIiEcviEv(%struct.S.0* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]])
// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CALL]], [[CALL2]]
// CHECK2-NEXT:    call void @_ZN1SIiEC1Ei(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]], i32 [[ADD]])
// CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]] to i8*
// CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[REF_TMP]] to i8*
// CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP13]], i8* align 4 [[TMP14]], i64 4, i1 false)
// CHECK2-NEXT:    call void @_ZN1SIiED1Ev(%struct.S.0* nonnull align 4 dereferenceable(4) [[REF_TMP]]) #[[ATTR5]]
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
// CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
// CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
// CHECK2:       omp.arraycpy.done3:
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP0:%.*]] = load volatile double, double* @g, align 8
// CHECK2-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i32
// CHECK2-NEXT:    store i32 [[CONV]], i32* [[F]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 [[A:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], %struct.S.0* [[THIS1]], i32 0, i32 0
// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
// CHECK2-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to double
// CHECK2-NEXT:    [[TMP1:%.*]] = load volatile double, double* @g, align 8
// CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], [[TMP1]]
// CHECK2-NEXT:    [[CONV2:%.*]] = fptosi double [[ADD]] to i32
// CHECK2-NEXT:    store i32 [[CONV2]], i32* [[F]], align 4
// CHECK2-NEXT:    ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
// CHECK2-SAME: (%struct.S.0* nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] align 2 {
// CHECK2-NEXT:  entry:
// CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S.0*, align 8
// CHECK2-NEXT:    store %struct.S.0* [[THIS]], %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S.0*, %struct.S.0** [[THIS_ADDR]], align 8
// CHECK2-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, i32* [[RETVAL]], align 4
// CHECK3-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 1 dereferenceable(1) [[REF_TMP]])
// CHECK3-NEXT:    ret i32 0
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[G1:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca double*, align 8
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
// CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[_TMP7:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[ATOMIC_TEMP11:%.*]] = alloca double, align 8
// CHECK3-NEXT:    [[_TMP12:%.*]] = alloca double, align 8
// CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load double*, double** @g1, align 8
// CHECK3-NEXT:    store double* [[TMP0]], double** [[TMP]], align 8
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT:    store double 0.000000e+00, double* [[G]], align 8
// CHECK3-NEXT:    [[TMP1:%.*]] = load double*, double** @g1, align 8
// CHECK3-NEXT:    store double 0.000000e+00, double* [[G1]], align 8
// CHECK3-NEXT:    store double* [[G1]], double** [[_TMP2]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
// CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3:       cond.true:
// CHECK3-NEXT:    br label [[COND_END:%.*]]
// CHECK3:       cond.false:
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    br label [[COND_END]]
// CHECK3:       cond.end:
// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT:    store double 1.000000e+00, double* [[G]], align 8
// CHECK3-NEXT:    [[TMP10:%.*]] = load double*, double** [[_TMP2]], align 8
// CHECK3-NEXT:    store volatile double 1.000000e+00, double* [[TMP10]], align 8
// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0
// CHECK3-NEXT:    store double* [[G]], double** [[TMP11]], align 8
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 1
// CHECK3-NEXT:    [[TMP13:%.*]] = load double*, double** [[_TMP2]], align 8
// CHECK3-NEXT:    store double* [[TMP13]], double** [[TMP12]], align 8
// CHECK3-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 8 dereferenceable(16) [[REF_TMP]])
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3:       omp.body.continue:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3:       omp.inner.for.inc:
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
// CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3:       omp.loop.exit:
// CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK3-NEXT:    [[TMP16:%.*]] = bitcast double* [[G]] to i8*
// CHECK3-NEXT:    store i8* [[TMP16]], i8** [[TMP15]], align 8
// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK3-NEXT:    [[TMP18:%.*]] = bitcast double* [[G1]] to i8*
// CHECK3-NEXT:    store i8* [[TMP18]], i8** [[TMP17]], align 8
// CHECK3-NEXT:    [[TMP19:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 2, i64 16, i8* [[TMP19]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK3-NEXT:    switch i32 [[TMP20]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK3-NEXT:    ]
// CHECK3:       .omp.reduction.case1:
// CHECK3-NEXT:    [[TMP21:%.*]] = load double, double* @g, align 8
// CHECK3-NEXT:    [[TMP22:%.*]] = load double, double* [[G]], align 8
// CHECK3-NEXT:    [[ADD5:%.*]] = fadd double [[TMP21]], [[TMP22]]
// CHECK3-NEXT:    store double [[ADD5]], double* @g, align 8
// CHECK3-NEXT:    [[TMP23:%.*]] = load double, double* [[TMP1]], align 8
// CHECK3-NEXT:    [[TMP24:%.*]] = load double, double* [[G1]], align 8
// CHECK3-NEXT:    [[ADD6:%.*]] = fadd double [[TMP23]], [[TMP24]]
// CHECK3-NEXT:    store double [[ADD6]], double* [[TMP1]], align 8
// CHECK3-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK3:       .omp.reduction.case2:
// CHECK3-NEXT:    [[TMP25:%.*]] = load double, double* [[G]], align 8
// CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i64, i64* bitcast (double* @g to i64*) monotonic, align 8
// CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
// CHECK3:       atomic_cont:
// CHECK3-NEXT:    [[TMP26:%.*]] = phi i64 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP33:%.*]], [[ATOMIC_CONT]] ]
// CHECK3-NEXT:    [[TMP27:%.*]] = bitcast double* [[ATOMIC_TEMP]] to i64*
// CHECK3-NEXT:    [[TMP28:%.*]] = bitcast i64 [[TMP26]] to double
// CHECK3-NEXT:    store double [[TMP28]], double* [[_TMP7]], align 8
// CHECK3-NEXT:    [[TMP29:%.*]] = load double, double* [[_TMP7]], align 8
// CHECK3-NEXT:    [[TMP30:%.*]] = load double, double* [[G]], align 8
// CHECK3-NEXT:    [[ADD8:%.*]] = fadd double [[TMP29]], [[TMP30]]
// CHECK3-NEXT:    store double [[ADD8]], double* [[ATOMIC_TEMP]], align 8
// CHECK3-NEXT:    [[TMP31:%.*]] = load i64, i64* [[TMP27]], align 8
// CHECK3-NEXT:    [[TMP32:%.*]] = cmpxchg i64* bitcast (double* @g to i64*), i64 [[TMP26]], i64 [[TMP31]] monotonic monotonic, align 8
// CHECK3-NEXT:    [[TMP33]] = extractvalue { i64, i1 } [[TMP32]], 0
// CHECK3-NEXT:    [[TMP34:%.*]] = extractvalue { i64, i1 } [[TMP32]], 1
// CHECK3-NEXT:    br i1 [[TMP34]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
// CHECK3:       atomic_exit:
// CHECK3-NEXT:    [[TMP35:%.*]] = load double, double* [[G1]], align 8
// CHECK3-NEXT:    [[TMP36:%.*]] = bitcast double* [[TMP1]] to i64*
// CHECK3-NEXT:    [[ATOMIC_LOAD9:%.*]] = load atomic i64, i64* [[TMP36]] monotonic, align 8
// CHECK3-NEXT:    br label [[ATOMIC_CONT10:%.*]]
// CHECK3:       atomic_cont10:
// CHECK3-NEXT:    [[TMP37:%.*]] = phi i64 [ [[ATOMIC_LOAD9]], [[ATOMIC_EXIT]] ], [ [[TMP45:%.*]], [[ATOMIC_CONT10]] ]
// CHECK3-NEXT:    [[TMP38:%.*]] = bitcast double* [[ATOMIC_TEMP11]] to i64*
// CHECK3-NEXT:    [[TMP39:%.*]] = bitcast i64 [[TMP37]] to double
// CHECK3-NEXT:    store double [[TMP39]], double* [[_TMP12]], align 8
// CHECK3-NEXT:    [[TMP40:%.*]] = load double, double* [[_TMP12]], align 8
// CHECK3-NEXT:    [[TMP41:%.*]] = load double, double* [[G1]], align 8
// CHECK3-NEXT:    [[ADD13:%.*]] = fadd double [[TMP40]], [[TMP41]]
// CHECK3-NEXT:    store double [[ADD13]], double* [[ATOMIC_TEMP11]], align 8
// CHECK3-NEXT:    [[TMP42:%.*]] = load i64, i64* [[TMP38]], align 8
// CHECK3-NEXT:    [[TMP43:%.*]] = bitcast double* [[TMP1]] to i64*
// CHECK3-NEXT:    [[TMP44:%.*]] = cmpxchg i64* [[TMP43]], i64 [[TMP37]], i64 [[TMP42]] monotonic monotonic, align 8
// CHECK3-NEXT:    [[TMP45]] = extractvalue { i64, i1 } [[TMP44]], 0
// CHECK3-NEXT:    [[TMP46:%.*]] = extractvalue { i64, i1 } [[TMP44]], 1
// CHECK3-NEXT:    br i1 [[TMP46]], label [[ATOMIC_EXIT14:%.*]], label [[ATOMIC_CONT10]]
// CHECK3:       atomic_exit14:
// CHECK3-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK3:       .omp.reduction.default:
// CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
// CHECK3-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK3-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK3-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK3-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK3-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to double*
// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK3-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to double*
// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
// CHECK3-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to double*
// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK3-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to double*
// CHECK3-NEXT:    [[TMP18:%.*]] = load double, double* [[TMP11]], align 8
// CHECK3-NEXT:    [[TMP19:%.*]] = load double, double* [[TMP8]], align 8
// CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[TMP18]], [[TMP19]]
// CHECK3-NEXT:    store double [[ADD]], double* [[TMP11]], align 8
// CHECK3-NEXT:    [[TMP20:%.*]] = load double, double* [[TMP17]], align 8
// CHECK3-NEXT:    [[TMP21:%.*]] = load double, double* [[TMP14]], align 8
// CHECK3-NEXT:    [[ADD2:%.*]] = fadd double [[TMP20]], [[TMP21]]
// CHECK3-NEXT:    store double [[ADD2]], double* [[TMP17]], align 8
// CHECK3-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@main
// CHECK4-SAME: () #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
// CHECK4-NEXT:    [[TMP0:%.*]] = load i8*, i8** getelementptr inbounds ([[STRUCT___BLOCK_LITERAL_GENERIC:%.*]], %struct.__block_literal_generic* bitcast ({ i8**, i32, i32, i8*, %struct.__block_descriptor* }* @__block_literal_global to %struct.__block_literal_generic*), i32 0, i32 3), align 8
// CHECK4-NEXT:    [[TMP1:%.*]] = bitcast i8* [[TMP0]] to void (i8*)*
// CHECK4-NEXT:    call void [[TMP1]](i8* bitcast ({ i8**, i32, i32, i8*, %struct.__block_descriptor* }* @__block_literal_global to i8*))
// CHECK4-NEXT:    ret i32 0
//
//
// CHECK4-LABEL: define {{[^@]+}}@__main_block_invoke
// CHECK4-SAME: (i8* [[DOTBLOCK_DESCRIPTOR:%.*]]) #[[ATTR2:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTBLOCK_DESCRIPTOR_ADDR:%.*]] = alloca i8*, align 8
// CHECK4-NEXT:    [[BLOCK_ADDR:%.*]] = alloca <{ i8*, i32, i32, i8*, %struct.__block_descriptor* }>*, align 8
// CHECK4-NEXT:    store i8* [[DOTBLOCK_DESCRIPTOR]], i8** [[DOTBLOCK_DESCRIPTOR_ADDR]], align 8
// CHECK4-NEXT:    [[BLOCK:%.*]] = bitcast i8* [[DOTBLOCK_DESCRIPTOR]] to <{ i8*, i32, i32, i8*, %struct.__block_descriptor* }>*
// CHECK4-NEXT:    store <{ i8*, i32, i32, i8*, %struct.__block_descriptor* }>* [[BLOCK]], <{ i8*, i32, i32, i8*, %struct.__block_descriptor* }>** [[BLOCK_ADDR]], align 8
// CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT:    [[TMP:%.*]] = alloca double*, align 8
// CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[G:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[G1:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[_TMP2:%.*]] = alloca double*, align 8
// CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT:    [[BLOCK:%.*]] = alloca <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, align 8
// CHECK4-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
// CHECK4-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[_TMP8:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[ATOMIC_TEMP12:%.*]] = alloca double, align 8
// CHECK4-NEXT:    [[_TMP13:%.*]] = alloca double, align 8
// CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP0:%.*]] = load double*, double** @g1, align 8
// CHECK4-NEXT:    store double* [[TMP0]], double** [[TMP]], align 8
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT:    store double 0.000000e+00, double* [[G]], align 8
// CHECK4-NEXT:    [[TMP1:%.*]] = load double*, double** @g1, align 8
// CHECK4-NEXT:    store double 0.000000e+00, double* [[G1]], align 8
// CHECK4-NEXT:    store double* [[G1]], double** [[_TMP2]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
// CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4:       cond.true:
// CHECK4-NEXT:    br label [[COND_END:%.*]]
// CHECK4:       cond.false:
// CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    br label [[COND_END]]
// CHECK4:       cond.end:
// CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4:       omp.inner.for.cond:
// CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK4-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4:       omp.inner.for.body:
// CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
// CHECK4-NEXT:    store double 1.000000e+00, double* [[G]], align 8
// CHECK4-NEXT:    [[TMP10:%.*]] = load double*, double** [[_TMP2]], align 8
// CHECK4-NEXT:    store volatile double 1.000000e+00, double* [[TMP10]], align 8
// CHECK4-NEXT:    [[BLOCK_ISA:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 0
// CHECK4-NEXT:    store i8* bitcast (i8** @_NSConcreteStackBlock to i8*), i8** [[BLOCK_ISA]], align 8
// CHECK4-NEXT:    [[BLOCK_FLAGS:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 1
// CHECK4-NEXT:    store i32 1073741824, i32* [[BLOCK_FLAGS]], align 8
// CHECK4-NEXT:    [[BLOCK_RESERVED:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 2
// CHECK4-NEXT:    store i32 0, i32* [[BLOCK_RESERVED]], align 4
// CHECK4-NEXT:    [[BLOCK_INVOKE:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 3
// CHECK4-NEXT:    store i8* bitcast (void (i8*)* @g1_block_invoke to i8*), i8** [[BLOCK_INVOKE]], align 8
// CHECK4-NEXT:    [[BLOCK_DESCRIPTOR:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 4
// CHECK4-NEXT:    store %struct.__block_descriptor* bitcast ({ i64, i64, i8*, i8* }* @__block_descriptor_tmp.1 to %struct.__block_descriptor*), %struct.__block_descriptor** [[BLOCK_DESCRIPTOR]], align 8
// CHECK4-NEXT:    [[BLOCK_CAPTURED:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 5
// CHECK4-NEXT:    [[TMP11:%.*]] = load volatile double, double* [[G]], align 8
// CHECK4-NEXT:    store volatile double [[TMP11]], double* [[BLOCK_CAPTURED]], align 8
// CHECK4-NEXT:    [[BLOCK_CAPTURED4:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP12:%.*]] = load double*, double** [[_TMP2]], align 8
// CHECK4-NEXT:    store double* [[TMP12]], double** [[BLOCK_CAPTURED4]], align 8
// CHECK4-NEXT:    [[TMP13:%.*]] = bitcast <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]] to void ()*
// CHECK4-NEXT:    [[BLOCK_LITERAL:%.*]] = bitcast void ()* [[TMP13]] to %struct.__block_literal_generic*
// CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___BLOCK_LITERAL_GENERIC:%.*]], %struct.__block_literal_generic* [[BLOCK_LITERAL]], i32 0, i32 3
// CHECK4-NEXT:    [[TMP15:%.*]] = bitcast %struct.__block_literal_generic* [[BLOCK_LITERAL]] to i8*
// CHECK4-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP14]], align 8
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to void (i8*)*
// CHECK4-NEXT:    call void [[TMP17]](i8* [[TMP15]])
// CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK4:       omp.body.continue:
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4:       omp.inner.for.inc:
// CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP18]], 1
// CHECK4-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
// CHECK4:       omp.inner.for.end:
// CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4:       omp.loop.exit:
// CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
// CHECK4-NEXT:    [[TMP20:%.*]] = bitcast double* [[G]] to i8*
// CHECK4-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
// CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
// CHECK4-NEXT:    [[TMP22:%.*]] = bitcast double* [[G1]] to i8*
// CHECK4-NEXT:    store i8* [[TMP22]], i8** [[TMP21]], align 8
// CHECK4-NEXT:    [[TMP23:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
// CHECK4-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 2, i64 16, i8* [[TMP23]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK4-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
// CHECK4-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
// CHECK4-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
// CHECK4-NEXT:    ]
// CHECK4:       .omp.reduction.case1:
// CHECK4-NEXT:    [[TMP25:%.*]] = load double, double* @g, align 8
// CHECK4-NEXT:    [[TMP26:%.*]] = load double, double* [[G]], align 8
// CHECK4-NEXT:    [[ADD6:%.*]] = fadd double [[TMP25]], [[TMP26]]
// CHECK4-NEXT:    store double [[ADD6]], double* @g, align 8
// CHECK4-NEXT:    [[TMP27:%.*]] = load double, double* [[TMP1]], align 8
// CHECK4-NEXT:    [[TMP28:%.*]] = load double, double* [[G1]], align 8
// CHECK4-NEXT:    [[ADD7:%.*]] = fadd double [[TMP27]], [[TMP28]]
// CHECK4-NEXT:    store double [[ADD7]], double* [[TMP1]], align 8
// CHECK4-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK4-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK4:       .omp.reduction.case2:
// CHECK4-NEXT:    [[TMP29:%.*]] = load double, double* [[G]], align 8
// CHECK4-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i64, i64* bitcast (double* @g to i64*) monotonic, align 8
// CHECK4-NEXT:    br label [[ATOMIC_CONT:%.*]]
// CHECK4:       atomic_cont:
// CHECK4-NEXT:    [[TMP30:%.*]] = phi i64 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP37:%.*]], [[ATOMIC_CONT]] ]
// CHECK4-NEXT:    [[TMP31:%.*]] = bitcast double* [[ATOMIC_TEMP]] to i64*
// CHECK4-NEXT:    [[TMP32:%.*]] = bitcast i64 [[TMP30]] to double
// CHECK4-NEXT:    store double [[TMP32]], double* [[_TMP8]], align 8
// CHECK4-NEXT:    [[TMP33:%.*]] = load double, double* [[_TMP8]], align 8
// CHECK4-NEXT:    [[TMP34:%.*]] = load double, double* [[G]], align 8
// CHECK4-NEXT:    [[ADD9:%.*]] = fadd double [[TMP33]], [[TMP34]]
// CHECK4-NEXT:    store double [[ADD9]], double* [[ATOMIC_TEMP]], align 8
// CHECK4-NEXT:    [[TMP35:%.*]] = load i64, i64* [[TMP31]], align 8
// CHECK4-NEXT:    [[TMP36:%.*]] = cmpxchg i64* bitcast (double* @g to i64*), i64 [[TMP30]], i64 [[TMP35]] monotonic monotonic, align 8
// CHECK4-NEXT:    [[TMP37]] = extractvalue { i64, i1 } [[TMP36]], 0
// CHECK4-NEXT:    [[TMP38:%.*]] = extractvalue { i64, i1 } [[TMP36]], 1
// CHECK4-NEXT:    br i1 [[TMP38]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
// CHECK4:       atomic_exit:
// CHECK4-NEXT:    [[TMP39:%.*]] = load double, double* [[G1]], align 8
// CHECK4-NEXT:    [[TMP40:%.*]] = bitcast double* [[TMP1]] to i64*
// CHECK4-NEXT:    [[ATOMIC_LOAD10:%.*]] = load atomic i64, i64* [[TMP40]] monotonic, align 8
// CHECK4-NEXT:    br label [[ATOMIC_CONT11:%.*]]
// CHECK4:       atomic_cont11:
// CHECK4-NEXT:    [[TMP41:%.*]] = phi i64 [ [[ATOMIC_LOAD10]], [[ATOMIC_EXIT]] ], [ [[TMP49:%.*]], [[ATOMIC_CONT11]] ]
// CHECK4-NEXT:    [[TMP42:%.*]] = bitcast double* [[ATOMIC_TEMP12]] to i64*
// CHECK4-NEXT:    [[TMP43:%.*]] = bitcast i64 [[TMP41]] to double
// CHECK4-NEXT:    store double [[TMP43]], double* [[_TMP13]], align 8
// CHECK4-NEXT:    [[TMP44:%.*]] = load double, double* [[_TMP13]], align 8
// CHECK4-NEXT:    [[TMP45:%.*]] = load double, double* [[G1]], align 8
// CHECK4-NEXT:    [[ADD14:%.*]] = fadd double [[TMP44]], [[TMP45]]
// CHECK4-NEXT:    store double [[ADD14]], double* [[ATOMIC_TEMP12]], align 8
// CHECK4-NEXT:    [[TMP46:%.*]] = load i64, i64* [[TMP42]], align 8
// CHECK4-NEXT:    [[TMP47:%.*]] = bitcast double* [[TMP1]] to i64*
// CHECK4-NEXT:    [[TMP48:%.*]] = cmpxchg i64* [[TMP47]], i64 [[TMP41]], i64 [[TMP46]] monotonic monotonic, align 8
// CHECK4-NEXT:    [[TMP49]] = extractvalue { i64, i1 } [[TMP48]], 0
// CHECK4-NEXT:    [[TMP50:%.*]] = extractvalue { i64, i1 } [[TMP48]], 1
// CHECK4-NEXT:    br i1 [[TMP50]], label [[ATOMIC_EXIT15:%.*]], label [[ATOMIC_CONT11]]
// CHECK4:       atomic_exit15:
// CHECK4-NEXT:    call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], [8 x i32]* @.gomp_critical_user_.reduction.var)
// CHECK4-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
// CHECK4:       .omp.reduction.default:
// CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@g1_block_invoke
// CHECK4-SAME: (i8* [[DOTBLOCK_DESCRIPTOR:%.*]]) #[[ATTR2]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTBLOCK_DESCRIPTOR_ADDR:%.*]] = alloca i8*, align 8
// CHECK4-NEXT:    [[BLOCK_ADDR:%.*]] = alloca <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>*, align 8
// CHECK4-NEXT:    store i8* [[DOTBLOCK_DESCRIPTOR]], i8** [[DOTBLOCK_DESCRIPTOR_ADDR]], align 8
// CHECK4-NEXT:    [[BLOCK:%.*]] = bitcast i8* [[DOTBLOCK_DESCRIPTOR]] to <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>*
// CHECK4-NEXT:    store <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>** [[BLOCK_ADDR]], align 8
// CHECK4-NEXT:    [[BLOCK_CAPTURE_ADDR:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 5
// CHECK4-NEXT:    store double 2.000000e+00, double* [[BLOCK_CAPTURE_ADDR]], align 8
// CHECK4-NEXT:    [[BLOCK_CAPTURE_ADDR1:%.*]] = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>, <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, double, double* }>* [[BLOCK]], i32 0, i32 6
// CHECK4-NEXT:    [[TMP0:%.*]] = load double*, double** [[BLOCK_CAPTURE_ADDR1]], align 8
// CHECK4-NEXT:    store double 2.000000e+00, double* [[TMP0]], align 8
// CHECK4-NEXT:    ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
// CHECK4-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
// CHECK4-NEXT:  entry:
// CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
// CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
// CHECK4-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
// CHECK4-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
// CHECK4-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
// CHECK4-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
// CHECK4-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
// CHECK4-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
// CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
// CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
// CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to double*
// CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
// CHECK4-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
// CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to double*
// CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
// CHECK4-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
// CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8* [[TMP13]] to double*
// CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
// CHECK4-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
// CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8* [[TMP16]] to double*
// CHECK4-NEXT:    [[TMP18:%.*]] = load double, double* [[TMP11]], align 8
// CHECK4-NEXT:    [[TMP19:%.*]] = load double, double* [[TMP8]], align 8
// CHECK4-NEXT:    [[ADD:%.*]] = fadd double [[TMP18]], [[TMP19]]
// CHECK4-NEXT:    store double [[ADD]], double* [[TMP11]], align 8
// CHECK4-NEXT:    [[TMP20:%.*]] = load double, double* [[TMP17]], align 8
// CHECK4-NEXT:    [[TMP21:%.*]] = load double, double* [[TMP14]], align 8
// CHECK4-NEXT:    [[ADD2:%.*]] = fadd double [[TMP20]], [[TMP21]]
// CHECK4-NEXT:    store double [[ADD2]], double* [[TMP17]], align 8
// CHECK4-NEXT:    ret void
//